import math import os import timeit from functools import lru_cache from logging import Formatter, INFO, StreamHandler,FileHandler, getLogger import cv2 import numpy as np from numpy.linalg import _umath_linalg from EyeTrackApp.utils.img_utils import safe_crop from EyeTrackApp.utils.misc_utils import clamp from EyeTrackApp.utils.time_utils import FPSResult, TimeitResult, format_time # from line_profiler_pycharm import profile this_file_basename = os.path.basename(__file__) this_file_name = this_file_basename.replace(".py", "") alg_ver = "230314-1" # Do not change it. ############################## # These can be changed old_mode = False save_logfile = False # This setting is disabled when imshow_enable or save_video is true imshow_enable = False save_video = False loop_num = 1 if imshow_enable or save_video else 100 input_video_path = "Pro_demo2.mp4" output_video_path = f'./{this_file_name}_{alg_ver}_new.mp4' if not old_mode else f'./{this_file_name}_{alg_ver}_old.mp4' logfilename = f'./{this_file_name}_{alg_ver}_new.log' if not old_mode else f'./{this_file_name}_old.log' print_enable = False # I don't recommend changing to True. # RANSAC thresh_add = 10 # calc_print_enable = True skip_autoradius = False skip_blink_detect = False ############################## ############################## # Do not change these. imsave_flg = imshow_enable or save_video # cache param lru_maxsize_vvs = 16 lru_maxsize_vs = 64 lru_maxsize_s = 128 # CV param default_radius = 20 auto_radius_range = (default_radius - 10, default_radius + 10) # (10,30) auto_radius_step = 1 blink_init_frames = 60 * 3 # 60fps*3sec,Number of blink statistical frames # step==(x,y) default_step = (5, 5) # bigger the steps,lower the processing time! ofc acc also takes an impact logger = getLogger(__name__) logger.setLevel(INFO) formatter = Formatter('%(message)s') handler = StreamHandler() handler.setLevel(INFO) handler.setFormatter(formatter) logger.addHandler(handler) if save_logfile and not imsave_flg: handler = FileHandler(logfilename, encoding="utf8", mode="w") handler.setLevel(INFO) handler.setFormatter(formatter) logger.addHandler(handler) else: save_logfile = False video_wr = cv2.VideoWriter(output_video_path, cv2.VideoWriter_fourcc(*"x264"), 60.0, (200, 150)) if save_video else None ############################## class AutoRadiusCalc(object): def __init__(self): self.response_list = [] self.radius_cand_list = [] self.adj_comp_flag = False self.radius_middle_index = None self.left_item = None self.right_item = None self.left_index = None self.right_index = None def get_radius(self): prev_res_len = len(self.response_list) # adjustment of radius if prev_res_len == 1: # len==1==response_list==[default_radius] self.adj_comp_flag = False return auto_radius_range[0] elif prev_res_len == 2: # len==2==response_list==[default_radius, auto_radius_range[0]] self.adj_comp_flag = False return auto_radius_range[1] elif prev_res_len == 3: # len==3==response_list==[default_radius,auto_radius_range[0],auto_radius_range[1]] if self.response_list[1][1] < self.response_list[2][1]: self.left_item = self.response_list[1] self.right_item = self.response_list[0] else: self.left_item = self.response_list[0] self.right_item = self.response_list[2] self.radius_cand_list = [i for i in range(self.left_item[0], self.right_item[0] + auto_radius_step, auto_radius_step)] self.left_index = 0 self.right_index = len(self.radius_cand_list) - 1 self.radius_middle_index = (self.left_index + self.right_index) // 2 self.adj_comp_flag = False return self.radius_cand_list[self.radius_middle_index] else: if self.left_index <= self.right_index and self.left_index != self.radius_middle_index: if (self.left_item[1] + self.response_list[-1][1]) < (self.right_item[1] + self.response_list[-1][1]): self.right_item = self.response_list[-1] self.right_index = self.radius_middle_index - 1 self.radius_middle_index = (self.left_index + self.right_index) // 2 self.adj_comp_flag = False return self.radius_cand_list[self.radius_middle_index] if (self.left_item[1] + self.response_list[-1][1]) > (self.right_item[1] + self.response_list[-1][1]): self.left_item = self.response_list[-1] self.left_index = self.radius_middle_index + 1 self.radius_middle_index = (self.left_index + self.right_index) // 2 self.adj_comp_flag = False return self.radius_cand_list[self.radius_middle_index] self.adj_comp_flag = True return self.radius_cand_list[self.radius_middle_index] def get_radius_base(self): """ Use it when the new version doesn't work well. :return: """ prev_res_len = len(self.response_list) # adjustment of radius if prev_res_len == 1: # len==1==response_list==[default_radius] self.adj_comp_flag = False return auto_radius_range[0] elif prev_res_len == 2: # len==2==response_list==[default_radius, auto_radius_range[0]] self.adj_comp_flag = False return auto_radius_range[1] elif prev_res_len == 3: # len==3==response_list==[default_radius,auto_radius_range[0],auto_radius_range[1]] sort_res = sorted(self.response_list, key=lambda x: x[1])[0] # Extract the radius with the lowest response value if sort_res[0] == default_radius: # If the default value is best, change now_mode to init after setting radius to the default value. self.adj_comp_flag = True return default_radius elif sort_res[0] == auto_radius_range[0]: self.radius_cand_list = [i for i in range(auto_radius_range[0], default_radius, auto_radius_step)][1:] self.adj_comp_flag = False return self.radius_cand_list.pop() else: self.radius_cand_list = [i for i in range(default_radius, auto_radius_range[1], auto_radius_step)][1:] self.adj_comp_flag = False return self.radius_cand_list.pop() else: # Try the contents of the radius_cand_list in order until the radius_cand_list runs out # Better make it a binary search. if len(self.radius_cand_list) == 0: sort_res = sorted(self.response_list, key=lambda x: x[1])[0] self.adj_comp_flag = True return sort_res[0] else: self.adj_comp_flag = False return self.radius_cand_list.pop() def add_response(self, radius, response): self.response_list.append((radius, response)) return None class BlinkDetector(object): def __init__(self): self.response_list = [] self.response_max = None self.enable_detect_flg = False self.quartile_1 = None def calc_thresh(self): # Calculate response_max by computing interquartile range, IQR # self.response_listo = np.array(self.response_listo) # 25%,75% # This value may need to be adjusted depending on the environment. # quartile_1, quartile_3 = np.percentile(self.response_listo, [25, 75]) # iqr = quartile_3 - quartile_1 # self.response_maxo = quartile_3 + (iqr * 1.5) # quartile_1, quartile_3 = np.percentile(self.response_list, [25, 75]) # or quartile_1, quartile_3 = np.percentile(np.array(self.response_list), [25, 75]) self.quartile_1 = quartile_1 iqr = quartile_3 - quartile_1 # response_min = quartile_1 - (iqr * 1.5) self.response_max = float(quartile_3 + (iqr * 1.5)) # or # self.response_max = quartile_3 + (iqr * 1.5) self.enable_detect_flg = True return None def detect(self, now_response): return now_response > self.response_max def add_response(self, response): self.response_list.append(response) return None def response_len(self): return len(self.response_list) def ellipse_model(data, y, f): """ There is no need to make this process a function, since making the process a function will slow it down a little by calling it. The results may be slightly different from the lambda version due to calculation errors derived from float types, but the calculation results are virtually the same. a = 1.0,b = P[0],c = P[1],d = P[2],e = P[3],f = P[4] :param data: :param y: np.c_[d, e, a, c, b] :param f: f == P[4, 0] :return: this_return == np.array([ellipse_model(x, y) for (x, y) in data ]) """ return data.dot(y) + f def fit_rotated_ellipse_ransac_old(data: np.ndarray, rng: np.random.Generator, iter=100, sample_num=10, offset=80): # before changing these values, please read up on the ransac algorithm # However if you want to change any value just know that higher iterations will make processing frames slower effective_sample = None # The array contents do not change during the loop, so only one call is needed. # They say len is faster than shape. # Reference url: https://stackoverflow.com/questions/35547853/what-is-faster-python3s-len-or-numpys-shape len_data = len(data) if len_data < sample_num: return None # Type of calculation result ret_dtype = np.float64 # Sorts a random number array of size (iter,len_data). After sorting, returns the index of sample_num random numbers before sorting. # If the array size is less than about 100, this is faster than rng.choice. rng_sample = rng.random((iter, len_data)).argsort()[:, :sample_num] # or # I don't see any advantage to doing this. # rng_sample = np.asarray(rng.random((iter, len_data)).argsort()[:, :sample_num], dtype=np.int32) # I don't think it looks beautiful. # x,y,x**2,y**2,x*y,1,-1*x**2 datamod = np.concatenate( [data, data ** 2, (data[:, 0] * data[:, 1])[:, np.newaxis], np.ones((len_data, 1), dtype=ret_dtype), (-1 * data[:, 0] ** 2)[:, np.newaxis]], axis=1, dtype=ret_dtype) datamod_slim = np.array(datamod[:, :5], dtype=ret_dtype) datamod_rng = datamod[rng_sample] datamod_rng6 = datamod_rng[:, :, 6] datamod_rng_swap = datamod_rng[:, :, [4, 3, 0, 1, 5]] datamod_rng_swap_trans = datamod_rng_swap.transpose((0, 2, 1)) # These two lines are one of the bottlenecks datamod_rng_5x5 = np.matmul(datamod_rng_swap_trans, datamod_rng_swap) datamod_rng_p5smp = np.matmul(np.linalg.inv(datamod_rng_5x5), datamod_rng_swap_trans) datamod_rng_p = np.matmul(datamod_rng_p5smp, datamod_rng6[:, :, np.newaxis]).reshape((-1, 5)) # I don't think it looks beautiful. ellipse_y_arr = np.asarray( [datamod_rng_p[:, 2], datamod_rng_p[:, 3], np.ones(len(datamod_rng_p)), datamod_rng_p[:, 1], datamod_rng_p[:, 0]], dtype=ret_dtype) ellipse_data_arr = ellipse_model(datamod_slim, ellipse_y_arr, np.asarray(datamod_rng_p[:, 4])).transpose((1, 0)) ellipse_data_abs = np.abs(ellipse_data_arr) ellipse_data_index = np.argmax(np.sum(ellipse_data_abs < offset, axis=1), axis=0) effective_data_arr = ellipse_data_arr[ellipse_data_index] effective_sample_p_arr = datamod_rng_p[ellipse_data_index] return fit_rotated_ellipse_old(effective_data_arr, effective_sample_p_arr) # @profile def fit_rotated_ellipse_old(data, P): a = 1.0 b = P[0] c = P[1] d = P[2] e = P[3] f = P[4] # The cost of trigonometric functions is high. theta = 0.5 * np.arctan(b / (a - c), dtype=np.float64) theta_sin = np.sin(theta, dtype=np.float64) theta_cos = np.cos(theta, dtype=np.float64) tc2 = theta_cos ** 2 ts2 = theta_sin ** 2 b_tcs = b * theta_cos * theta_sin # Do the calculation only once cxy = b ** 2 - 4 * a * c cx = (2 * c * d - b * e) / cxy cy = (2 * a * e - b * d) / cxy # I just want to clear things up around here. cu = a * cx ** 2 + b * cx * cy + c * cy ** 2 - f cu_r = np.array([(a * tc2 + b_tcs + c * ts2), (a * ts2 - b_tcs + c * tc2)]) wh = np.sqrt(cu / cu_r) w, h = wh[0], wh[1] error_sum = np.sum(data) # print("fitting error = %.3f" % (error_sum)) return (cx, cy, w, h, theta) @lru_cache(maxsize=lru_maxsize_vs) def get_ransac_empty_array_lendata(len_data,iter_num, sample_num): # Function to reduce array allocation by providing an empty array first and recycling it with lru use_dtype=np.float64 datamod = np.empty((len_data, 7), dtype=use_dtype) # np.empty((len(data), 7), dtype=ret_dtype) datamod[:, 5] = 1 datamod_b=datamod[:, :5]#.T random_index_init_arr = np.empty((iter_num, len_data), dtype=np.uint16) random_index_init_arr[:, :] = np.arange(len_data, dtype=np.uint16) random_index = np.empty((iter_num, len_data), dtype=np.uint16) random_index_samplenum=random_index[:, :sample_num] ellipse_data_arr=np.empty((iter_num,len_data),dtype=use_dtype) th_abs=np.empty((iter_num,len_data),dtype=use_dtype) dm_data_view=datamod[:, :2]# = data dm_p2_view=datamod[:, 2:4]# = data * data dm_mul_view=datamod[:, 4]# = data[:, 0] * data[:, 1] dm_neg_view=datamod[:, 6]# = -datamod[:, 2] # return datamod,random_index_init_arr,random_index,ellipse_data_arr,th_abs return datamod,datamod_b,dm_data_view,dm_p2_view,dm_mul_view,dm_neg_view, random_index_init_arr, random_index,random_index_samplenum, ellipse_data_arr, th_abs @lru_cache(maxsize=lru_maxsize_s) def get_ransac_empty_array_iternum_samplenum(iter_num, sample_num,len_data): # Function to reduce array allocation by providing an empty array first and recycling it with lru use_dtype=np.float64 datamod_rng=np.empty((iter_num,sample_num,7),dtype=use_dtype) datamod_rng_swap = np.empty((iter_num, sample_num, 5), dtype=use_dtype) datamod_rng_swap_trans=datamod_rng_swap.transpose((0,2,1)) # datamod_rng_swap_trans = np.empty((iter_num, 5,sample_num), dtype=use_dtype) datamod_rng_5x5= np.empty((iter_num, 5,5), dtype=use_dtype) datamod_rng_p5smp = np.empty((iter_num, 5,sample_num), dtype=use_dtype) datamod_rng_p=np.empty((iter_num,5),dtype=use_dtype) datamod_rng_p_npaxis=datamod_rng_p[:,:,np.newaxis] ellipse_y_arr=np.empty((iter_num,5),dtype=use_dtype) ellipse_y_arr[:, 2] = 1 swap_index=np.array([4, 3, 0, 1, 5]) dm_brod=np.broadcast_to(datamod_rng_p[:, 4, np.newaxis], (iter_num, len_data)) dm_rng_six=datamod_rng[:, :, 6, np.newaxis] dm_rng_p_24_view= datamod_rng_p[:, 2:4] dm_rng_p_10_view= datamod_rng_p[:, 1::-1] el_y_arr_2_view=ellipse_y_arr[:, :2] el_y_arr_3_view=ellipse_y_arr[:, 3:] return datamod_rng,datamod_rng_swap,datamod_rng_swap_trans,datamod_rng_5x5,datamod_rng_p5smp,datamod_rng_p,datamod_rng_p_npaxis,ellipse_y_arr,swap_index,dm_brod,dm_rng_six,dm_rng_p_24_view,dm_rng_p_10_view,el_y_arr_2_view,el_y_arr_3_view # @profile def fit_rotated_ellipse_ransac(data: np.ndarray, sfc: np.random.Generator, iter_num=100, sample_num=10, offset=80 # 80.0, 10, 80 ): # before changing these values, please read up on the ransac algorithm # However if you want to change any value just know that higher iterations will make processing frames slower # The array contents do not change during the loop, so only one call is needed. # They say len is faster than shape. # Reference url: https://stackoverflow.com/questions/35547853/what-is-faster-python3s-len-or-numpys-shape len_data = len(data) if len_data < sample_num: return None # Type of calculation result # ret_dtype = np.float64 # todo:create view datamod_rng, datamod_rng_swap, datamod_rng_swap_trans, datamod_rng_5x5, datamod_rng_p5smp, datamod_rng_p,datamod_rng_p_npaxis, ellipse_y_arr,swap_index,dm_brod,dm_rng_six,dm_rng_p_24_view,dm_rng_p_10_view,el_y_arr_2_view,el_y_arr_3_view=get_ransac_empty_array_iternum_samplenum(iter_num,sample_num,len_data) datamod,datamod_b,dm_data_view,dm_p2_view,dm_mul_view,dm_neg_view,random_index_init_arr,random_index,random_index_samplenum,ellipse_data_arr,th_abs=get_ransac_empty_array_lendata(len_data,iter_num,sample_num) # Sorts a random number array of size (iter,len_data). After sorting, returns the index of sample_num random numbers before sorting. # If the array size is less than about 100, this is faster than rng.choice. # rng_sample = rng.random((iter_num, len_data),dtype=np.float32).argsort()[:, :sample_num]#out= # or # I don't see any advantage to doing this. # rng_sample = np.asarray(rng.random((iter, len_data)).argsort()[:, :sample_num], dtype=np.int32) # I don't think it looks beautiful. # x,y,x**2,y**2,x*y,1,-1*x**2 # datamod = np.concatenate( # [data, data ** 2, (data[:, 0] * data[:, 1])[:, np.newaxis], np.ones((len_data, 1), dtype=ret_dtype), # (-1 * data[:, 0] ** 2)[:, np.newaxis]], axis=1, # dtype=ret_dtype) # I don't think it looks beautiful. # x,y,x**2,y**2,x*y,1,-1*x**2 # data_squared = np.square(data) # datamod = np.empty((len(data), 7), dtype=ret_dtype) # np.empty((len(data), 7), dtype=ret_dtype) # datamod[:, :2] = data#[:] # # datamod[:, 2:4] = np.square(data) # or data**2 # # np.square(data,out=datamod[:, 2:4])#casting,dtype # datamod[:, 2:4] = data * data # datamod[:, 4] = data[:, 0] * data[:, 1] # # datamod[:, 4]=data.prod(axis=1,dtype=np.float64) # # datamod[:, 5] = 1 # datamod[:, 6] = -datamod[:, 2] # -1 * data[:, 0] ** 2# dm_data_view[:, :] = data#[:] dm_p2_view[:,:] = data * data dm_mul_view[:] = data[:, 0] * data[:, 1] dm_neg_view[:] = -dm_p2_view[:,0] # -1 * data[:, 0] ** 2# # datamod_slim = np.array(datamod[:, :5], dtype=ret_dtype) # # datamod_rng = datamod[rng_sample] # datamod_rng = rng.choice(datamod, (iter, sample_num), replace=True, shuffle=False) # datamod_rng = datamod[rng.choice(len_data, (iter,len_data),shuffle=False)[:,:sample_num]] # random_index = np.empty((iter_num, len_data), dtype=np.uint16) # random_index[:, :] = np.arange(len_data, dtype=np.uint16) # random_index_bro = np.broadcast_to(np.arange(len_data, dtype=np.uint16), (iter_num, len_data)) # or # random_index = np.zeros(100,dtype=np.uint16).reshape((-1,1))+ np.arange(len_data, dtype=np.uint16).reshape((1, len_data)) # sfc.permuted(random_index, axis=1, out=random_index) sfc.permuted(random_index_init_arr, axis=1, out=random_index) # random_index = sfc.permuted(np.broadcast_to(np.arange(len_data, dtype=np.uint16), (iter_num, len_data)),axis=1) # datamod_rng = datamod[random_index[:,:sample_num]]#take # np.take replaces a[ind,:] and is 3-4 times faster, https://gist.github.com/rossant/4645217 # datamod_rng = np.take(datamod,random_index[:,:sample_num],axis=0) # datamod_rng = datamod.take(random_index[:, :sample_num], axis=0, mode="clip")# iter_num,sample_num,7 # datamod.take(random_index[:, :sample_num], axis=0, mode="clip",out=datamod_rng) # iter_num,sample_num,7 datamod.take(random_index_samplenum, axis=0, mode="clip", out=datamod_rng) # datamod_rng = datamod[rng_sample]#out= # datamod_rng_swap = datamod_rng[:, :, [4, 3, 0, 1, 5]] # out= # iter_num,sample_num,5 # datamod_rng_swap[:,:,:] = datamod_rng[:, :, [4, 3, 0, 1, 5]] # out= # iter_num,sample_num, datamod_rng.take(swap_index, axis=2,mode="clip",out=datamod_rng_swap) # or # datamod_rng_swap = np.take(datamod_rng,[4, 3, 0, 1, 5],axis=2) # datamod_rng_swap_trans = datamod_rng_swap.transpose((0, 2, 1)) # out= # datamod_rng_swap_trans[:,:,:] = datamod_rng_swap.transpose((0, 2, 1))#.copy() # out= # # # These two lines are one of the bottlenecks # datamod_rng_5x5 = np.matmul(datamod_rng_swap_trans, datamod_rng_swap) # datamod_rng_p5smp = np.matmul(np.linalg.inv(datamod_rng_5x5), datamod_rng_swap_trans) # datamod_rng_p5smp = np.matmul(np.linalg.inv(np.matmul(datamod_rng_swap_trans, datamod_rng_swap)), datamod_rng_swap_trans) np.matmul(datamod_rng_swap_trans, datamod_rng_swap,out=datamod_rng_5x5) # I want to use cv2.mulTransposed, but for some reason the results are different and it can only use 1-channel arrays. # np.linalg.inv(datamod_rng_5x5) # datamod_rng_5x5[:,:,:]=np.linalg.inv(datamod_rng_5x5) # _umath_linalg.inv(datamod_rng_5x5,out=datamod_rng_5x5)# check error # https://github.com/bogovicj/JaneliaMLCourse/issues/1 # solve is slow # np.linalg.solve(np.matmul(datamod_rng_swap_trans, datamod_rng_swap), datamod_rng_swap_trans) _umath_linalg.inv(datamod_rng_5x5, signature='d->d', extobj=np.linalg.linalg.get_linalg_error_extobj(np.linalg.linalg._raise_linalgerror_singular),out=datamod_rng_5x5) np.matmul(datamod_rng_5x5, datamod_rng_swap_trans,out=datamod_rng_p5smp) # global ein_path # if ein_path is None: # ein_path = np.einsum_path("ijk,ijl->ikl", datamod_rng_swap, datamod_rng_swap, optimize='optimal')[0]#'optimal','greedy' # datamod_rng_p5smp = np.matmul(np.linalg.inv(np.einsum("ijk,ijl->ikl", datamod_rng_swap, datamod_rng_swap,casting="no",optimize=ein_path)), datamod_rng_swap_trans) # # np.einsum('ijk,ilk->ijl', dataswap_trans, dataswap) # # datamod_rng_p = np.matmul(datamod_rng_p5smp, datamod_rng[:, :, 6, np.newaxis]).reshape((-1, 5)) # out= # iter_num,5 # datamod_rng_p[:,:]=np.matmul(datamod_rng_p5smp, datamod_rng[:, :, 6, np.newaxis]).reshape((-1, 5)) # out= # iter_num,5 np.matmul(datamod_rng_p5smp, dm_rng_six,out=datamod_rng_p_npaxis) # # # I don't think it looks beautiful. # ellipse_y_arr = np.asarray( # [datamod_rng_p[:, 2], datamod_rng_p[:, 3], np.ones(iter_num,dtype=ret_dtype), datamod_rng_p[:, 1], datamod_rng_p[:, 0]], dtype=ret_dtype) # ellipse_y_arr = np.empty((iter_num, 5), dtype=ret_dtype) # ellipse_y_arr[0,:]=datamod_rng_p[:, 2] # ellipse_y_arr[1, :]=datamod_rng_p[:, 3] # ellipse_y_arr[:, :2] = datamod_rng_p[:, 2:4] # # ellipse_y_arr[:, 2] = 1 # ellipse_y_arr[:, 3:] = datamod_rng_p[:, 1::-1] el_y_arr_2_view[:,:]= dm_rng_p_24_view#datamod_rng_p[:, 2:4] el_y_arr_3_view[:,:] = dm_rng_p_10_view#datamod_rng_p[:, 1::-1] # ellipse_y_arr[:,3:]=datamod_rng_p[:,1] # ellipse_y_arr[4,:]=datamod_rng_p[:,0] # ellipse_data_arr = np.asarray(ellipse_model(datamod[:, :5], ellipse_y_arr.T, np.asarray(datamod_rng_p[:, 4])).transpose((1, 0))) # ellipse_data_arr = datamod[:, :5].dot(ellipse_y_arr.T)+np.asarray(datamod_rng_p[:, 4]) # ellipse_data_arr = np.matmul(ellipse_y_arr, datamod[:, :5].T) + np.asarray(datamod_rng_p[:, 4, np.newaxis])# iter_num,len_data # np.matmul(ellipse_y_arr, datamod_t,out=ellipse_data_arr) # ellipse_y_arr.dot(datamod[:, :5].T, out=ellipse_data_arr) # ellipse_data_arr+=datamod_rng_p[:, 4, np.newaxis]#np.asarray(datamod_rng_p[:, 4, np.newaxis]) # cv2.gemm is slower and for some reason the src3 argument for addition is not available cv2.gemm(ellipse_y_arr,datamod_b,1.0,dm_brod,1.0,dst=ellipse_data_arr,flags=cv2.GEMM_2_T) # ellipse_data_arr[:,:]=scipy.linalg.blas.dgemm(alpha=1.0, a=ellipse_y_arr, b=datamod_b, beta=1.0, c=np.broadcast_to(datamod_rng_p[:, 4, np.newaxis], (iter_num, len_data)),trans_b=True, overwrite_c=False) # ellipse_data_arr=ellipse_y_arr.dot(datamod[:, :5].T) + np.asarray(datamod_rng_p[:, 4, np.newaxis]) # ellipse_data_arr =ellipse_data_arr.transpose((1, 0)) # ellipse_data_arr = np.einsum("ij,kj->ki",np.asarray(datamod[:, :5]),ellipse_y_arr)+np.asarray(datamod_rng_p[:, 4,np.newaxis]) # Q, R = np.linalg.qr(datamod_rng_swap_trans) # # datarng_T = datamod_rng.transpose((0, 2, 1)) # Qtb = Q @ datamod_rng[:, :, 6].reshape((iter_num,5,5)) # p = np.linalg.solve(R, Qtb) # Q, R = np.linalg.qr(datamod_rng_swap)#, mode='raw') # Qtb = Q.transpose((0, 2, 1)) @ datamod_rng[:, :, 6, np.newaxis] # p = np.linalg.solve(R, Qtb.reshape((-1, 5))) # hoge=datamod_rng_swap.transpose((0, 2, 1))@datamod_rng_swap # inv_h=np.linalg.inv(hoge) # D, U = np.linalg.eigh(hoge) # Ap = (U * np.sqrt(D)).T # smp=np.linalg.solve(R, Q.transpose((0, 2, 1))) # smp= np.matmul(np.linalg.inv(R), Q.transpose((0, 2, 1))) # datamod_rng_p=np.matmul(smp, datamod_rng[:, :,6, np.newaxis]).reshape((-1, 5)) # ellipse_y_arr = np.asarray([datamod_rng_p[:,2], datamod_rng_p[:,3], np.ones(iter_num), datamod_rng_p[:,1], datamod_rng_p[:,0]], dtype=ret_dtype) # ellipse_data_arr = datamod[:, :5].dot(ellipse_y_arr) + datamod_rng_p[:,4] # ellipse_data_arr = ellipse_data_arr.transpose((1, 0)) # ellipse_data_abs = np.abs(ellipse_data_arr) # ellipse_data_abs = cv2.absdiff(ellipse_data_arr, 0) # ellipse_data_index = np.argmax(np.sum(ellipse_data_abs < offset, axis=1), axis=0) # ellipse_data_index = np.argmax(cv2.reduce(cv2.threshold(ellipse_data_abs, offset, 1, cv2.THRESH_BINARY_INV)[1], 1, cv2.REDUCE_SUM).reshape(-1), axis=0) # ellipse_data_index = cv2.reduceArgMax(cv2.reduce((ellipse_data_abs < offset)*1.0, 1, cv2.REDUCE_SUM),axis=0)[0,0] # ellipse_data_index = cv2.reduceArgMax(cv2.reduce(cv2.threshold(ellipse_data_abs, offset, 1, cv2.THRESH_BINARY_INV)[1], 1, cv2.REDUCE_SUM),axis=0)[0,0] # ellipse_data_index = np.einsum("ij->i", cv2.threshold(ellipse_data_abs, offset, 1, cv2.THRESH_BINARY_INV)[1]).argmax() # ellipse_data_index = \ # cv2.minMaxLoc(cv2.reduce(cv2.threshold(np.abs(ellipse_data_arr), offset, 1, cv2.THRESH_BINARY_INV)[1], 1, cv2.REDUCE_SUM))[3][1] np.abs(ellipse_data_arr,out=th_abs) cv2.threshold(th_abs, offset, 1.0, cv2.THRESH_BINARY_INV,dst=th_abs)#[1] ellipse_data_index = \ cv2.minMaxLoc(cv2.reduce(th_abs, 1, cv2.REDUCE_SUM))[3][1] # ellipse_data_index = np.linalg.norm(cv2.threshold(ellipse_data_abs, offset, 1, cv2.THRESH_BINARY_INV)[1],ord=0,axis=1).argmax() # if ellipse_data_index!=a: # print() # effective_data_arr = ellipse_data_arr[ellipse_data_index] # error_num = ellipse_data_arr[ellipse_data_index].sum() error_num = cv2.sumElems(ellipse_data_arr[ellipse_data_index])[0] effective_sample_p_arr = datamod_rng_p[ellipse_data_index].tolist() # if fit_rotated_ellipse(effective_data_arr.sum(), effective_sample_p_arr)!= fit_rotated_ellipse_base(effective_data_arr, effective_sample_p_arr): # print() return fit_rotated_ellipse(error_num, effective_sample_p_arr) # @profile def fit_rotated_ellipse(data, P): # a = 1.0 # # b, c, d, e, f = P # b, c, d, e, f = P[0], P[1], P[2], P[3], P[4] # # b = P[0] # # c = P[1] # # d = P[2] # # e = P[3] # # f = P[4] # # The cost of trigonometric functions is high. # theta = 0.5 * math.atan2(b, a-c)# math.atan(b / (a - c))# #np.arctan(b / (a - c), dtype=np.float64) # # theta_sin = np.sin(theta, dtype=np.float64) # # theta_cos = np.cos(theta, dtype=np.float64) # theta_sin, theta_cos = math.sin(theta), math.cos(theta) # tc2 = theta_cos ** 2 # ts2 = theta_sin ** 2 # b_tcs = b * theta_cos * theta_sin # # # Do the calculation only once # cxy = b ** 2 - 4 * a * c # cx = (2 * c * d - b * e) / cxy # cy = (2 * a * e - b * d) / cxy # # # I just want to clear things up around here. # cu = a * cx ** 2 + b * cx * cy + c * cy ** 2 - f # # cu_r = np.array([(a * tc2 + b_tcs + c * ts2), (a * ts2 - b_tcs + c * tc2)]) # # wh = np.sqrt(cu / cu_r) # # w = math.sqrt(cu/(a * tc2 + b_tcs + c * ts2)) # h = math.sqrt(cu/(a * ts2 - b_tcs + c * tc2)) a = 1.0 # b, c, d, e, f = P[0], P[1], P[2], P[3], P[4] b, c, d, e = P[0], P[1], P[2], P[3] theta = 0.5 * math.atan(b / (a - c)) # math.atan2(b, a - c) theta_sin, theta_cos = math.sin(theta), math.cos(theta) tc2 = theta_cos * theta_cos ts2 = theta_sin * theta_sin b_tcs = b * theta_cos * theta_sin cxy = b * b - 4 * a * c cx = (2 * c * d - b * e) / cxy cy = (2 * a * e - b * d) / cxy # cu = a * cx * cx + b * cx * cy + c * cy * cy - P[4]#f cu = c * cy * cy + cx * (a * cx + b * cy) - P[4] # here: https://stackoverflow.com/questions/327002/which-is-faster-in-python-x-5-or-math-sqrtx # and : https://gist.github.com/zed/783011 w = math.sqrt(cu / (a * tc2 + b_tcs + c * ts2)) h = math.sqrt(cu / (a * ts2 - b_tcs + c * tc2)) # error_sum = data.sum()#sum(data)#np.sum(data) # error_sum = data[0] + data[1] + data[2] + data[3] + data[4] + data[5] + data[6] + data[7] + data[8] + data[9] + data[10] error_sum = data # sum(data) # print("fitting error = %.3f" % (error_sum)) # cxy2 = P[0] * P[0] - 4 * a * P[1] # theta2 = 0.5 * math.atan(P[0] / (a - P[1])) # theta_sin2, theta_cos2 = math.sin(theta2), math.cos(theta2) # tc22 = theta_cos2 * theta_cos2 # ts22 = theta_sin2 * theta_sin2 # b_tcs2 = P[0] * theta_cos2 * theta_sin2 # cx2 = (2 * P[1] * P[2] - P[0] * P[3]) / cxy2 # cy2 = (2 * a * P[3] - P[0] * P[2]) / cxy2 # cu2 = P[1] * cy2 * cy2 + cx2 * (a * cx2 + P[0] * cy2) - P[4] # w2 = math.sqrt(cu2 / (a * tc22 + b_tcs2 + P[1] * ts22)) # h2 = math.sqrt(cu2 / (a * ts22 - b_tcs2 + P[1] * tc22)) return cx, cy, w, h, theta class CvParameters_old: # It may be a little slower because a dict named "self" is read for each function call. def __init__(self, radius, step): # self.prev_radius=radius self._radius = radius self.pad = 2 * radius # self.prev_step=step self._step = step self._hsf = HaarSurroundFeature_old(radius) def get_rpsh(self): return self._radius, self.pad, self._step, self._hsf # Essentially, the following would be preferable, but it would take twice as long to call. # return self.radius, self.pad, self.step, self.hsf @property def radius(self): return self._radius @radius.setter def radius(self, now_radius): # self.prev_radius=self._radius self._radius = now_radius self.pad = 2 * now_radius self.hsf = now_radius @property def step(self): return self._step @step.setter def step(self, now_step): # self.prev_step=self.step self._step = now_step @property def hsf(self): return self._hsf @hsf.setter def hsf(self, now_radius): self._hsf = HaarSurroundFeature_old(now_radius) class HaarSurroundFeature_old: def __init__(self, r_inner, r_outer=None, val=None): if r_outer is None: r_outer = r_inner * 3 # print(r_outer) r_inner2 = r_inner * r_inner count_inner = r_inner2 count_outer = r_outer * r_outer - r_inner2 if val is None: val_inner = 1.0 / r_inner2 val_outer = -val_inner * count_inner / count_outer else: val_inner = val[0] val_outer = val[1] self.val_in = np.array(val_inner, dtype=np.float64) self.val_out = np.array(val_outer, dtype=np.float64) self.r_in = r_inner self.r_out = r_outer def get_kernel(self): # Defined here, but not yet used? # Create a kernel filled with the value of self.val_out kernel = np.ones(shape=(2 * self.r_out - 1, 2 * self.r_out - 1), dtype=np.float64) * self.val_out # Set the values of the inner area of the kernel using array slicing start = (self.r_out - self.r_in) end = (self.r_out + self.r_in - 1) kernel[start:end, start:end] = self.val_in return kernel @lru_cache(maxsize=lru_maxsize_vvs) def get_hsf_empty_array_old(len_syx, frameint_x, frame_int_dtype, fcshape): # Function to reduce array allocation by providing an empty array first and recycling it with lru inner_sum = np.empty(len_syx, dtype=frame_int_dtype) outer_sum = np.empty(len_syx, dtype=frame_int_dtype) p_temp = np.empty((len_syx[0], frameint_x), dtype=frame_int_dtype) p00 = np.empty(len_syx, dtype=frame_int_dtype) p11 = np.empty(len_syx, dtype=frame_int_dtype) p01 = np.empty(len_syx, dtype=frame_int_dtype) p10 = np.empty(len_syx, dtype=frame_int_dtype) response_list = np.empty(len_syx, dtype=np.float64) frame_conv = np.zeros(shape=fcshape[0], dtype=np.uint8) frame_conv_stride = frame_conv[::fcshape[1], ::fcshape[2]] return (inner_sum, outer_sum), p_temp, (p00, p11, p01, p10), response_list, (frame_conv, frame_conv_stride) @lru_cache(maxsize=lru_maxsize_vs) def frameint_get_xy_step_old(imageshape, xysteps, pad, start_offset=None, end_offset=None): """ :param imageshape: (height(row),width(col)). row==y,cal==x :param xysteps: (x,y) :param pad: int :param start_offset: (x,y) or None :param end_offset: (x,y) or None :return: xy_np:tuple(x,y) """ row, col = imageshape row -= 1 col -= 1 x_step, y_step = xysteps # This is not beautiful. start_pad_x = start_pad_y = end_pad_x = end_pad_y = pad if start_offset is not None: start_pad_x += start_offset[0] start_pad_y += start_offset[1] if end_offset is not None: end_pad_x += end_offset[0] end_pad_y += end_offset[1] y_np = np.arange(start_pad_y, row - end_pad_y, y_step) x_np = np.arange(start_pad_x, col - end_pad_x, x_step) xy_np = (x_np, y_np) return xy_np # @profile def conv_int_old(frame_int, kernel, xy_step, padding, xy_steps_list): """ :param frame_int: :param kernel: hsf :param step: (x,y) :param padding: int :return: """ row, col = frame_int.shape row -= 1 col -= 1 x_step, y_step = xy_step # padding2 = 2 * padding f_shape = row - 2 * padding, col - 2 * padding r_in = kernel.r_in len_sx, len_sy = len(xy_steps_list[0]), len(xy_steps_list[1]) inout_sum, p_temp, p_list, response_list, frameconvlist = get_hsf_empty_array_old((len_sy, len_sx), col + 1, frame_int.dtype, (f_shape, y_step, x_step)) inner_sum, outer_sum = inout_sum p00, p11, p01, p10 = p_list frame_conv, frame_conv_stride = frameconvlist y_rin_m = xy_steps_list[1] - r_in x_rin_m = xy_steps_list[0] - r_in y_rin_p = xy_steps_list[1] + r_in x_rin_p = xy_steps_list[0] + r_in # xx==(y,x),m==MINUS,p==PLUS, ex: mm==(y-,x-) inarr_mm = frame_int[y_rin_m[0]:y_rin_m[-1] + 1:y_step, x_rin_m[0]:x_rin_m[-1] + 1:x_step] inarr_mp = frame_int[y_rin_m[0]:y_rin_m[-1] + 1:y_step, x_rin_p[0]:x_rin_p[-1] + 1:x_step] inarr_pm = frame_int[y_rin_p[0]:y_rin_p[-1] + 1:y_step, x_rin_m[0]:x_rin_m[-1] + 1:x_step] inarr_pp = frame_int[y_rin_p[0]:y_rin_p[-1] + 1:y_step, x_rin_p[0]:x_rin_p[-1] + 1:x_step] # == inarr_mm + inarr_pp - inarr_mp - inarr_pm inner_sum[:, :] = inarr_mm inner_sum += inarr_pp inner_sum -= inarr_mp inner_sum -= inarr_pm # Bottleneck here, I want to make it smarter. Someone do it. # (y,x) # p00=max(y_ro_m,0),max(x_ro_m,0) # p11=min(y_ro_p,ylim),min(x_ro_p,xlim) # p01=max(y_ro_m,0),min(x_ro_p,xlim) # p10=min(y_ro_p,ylim),max(x_ro_m,0) y_ro_m = xy_steps_list[1] - kernel.r_out x_ro_m = xy_steps_list[0] - kernel.r_out y_ro_p = xy_steps_list[1] + kernel.r_out x_ro_p = xy_steps_list[0] + kernel.r_out # p00 calc np.take(frame_int, y_ro_m, axis=0, mode="clip", out=p_temp) np.take(p_temp, x_ro_m, axis=1, mode="clip", out=p00) # p01 calc np.take(p_temp, x_ro_p, axis=1, mode="clip", out=p01) # p11 calc np.take(frame_int, y_ro_p, axis=0, mode="clip", out=p_temp) np.take(p_temp, x_ro_p, axis=1, mode="clip", out=p11) # p10 calc np.take(p_temp, x_ro_m, axis=1, mode="clip", out=p10) # the point is this # p00=np.take(np.take(frame_int, y_ro_m, axis=0, mode="clip"), x_ro_m, axis=1, mode="clip") # p11=np.take(np.take(frame_int, y_ro_p, axis=0, mode="clip"), x_ro_p, axis=1, mode="clip") # p01=np.take(np.take(frame_int, y_ro_m, axis=0, mode="clip"), x_ro_p, axis=1, mode="clip") # p10=np.take(np.take(frame_int, y_ro_p, axis=0, mode="clip"), x_ro_m, axis=1, mode="clip") outer_sum[:, :] = p00 + p11 - p01 - p10 - inner_sum np.multiply(kernel.val_in, inner_sum, dtype=np.float64, out=response_list) response_list += kernel.val_out * outer_sum # min_response, max_val, min_loc, max_loc = cv2.minMaxLoc(response_list) min_response, _, min_loc, _ = cv2.minMaxLoc(response_list) center = ((xy_steps_list[0][min_loc[0]] - padding), (xy_steps_list[1][min_loc[1]] - padding)) frame_conv_stride[:, :] = response_list # or # frame_conv_stride[:, :] = response_list.astype(np.uint8) return frame_conv, min_response, center class CvParameters_new: # It may be a little slower because a dict named "self" is read for each function call. def __init__(self, radius, step): # self.prev_radius=radius self._radius = radius self.pad = 2 * radius # self.prev_step=step self._step = step self._hsf = HaarSurroundFeature_new(radius) def get_rpsh(self): return self._radius, self.pad, self._step, self._hsf # Essentially, the following would be preferable, but it would take twice as long to call. # return self.radius, self.pad, self.step, self.hsf @property def radius(self): return self._radius @radius.setter def radius(self, now_radius): # self.prev_radius=self._radius self._radius = now_radius self.pad = 2 * now_radius self.hsf = now_radius @property def step(self): return self._step @step.setter def step(self, now_step): # self.prev_step=self.step self._step = now_step @property def hsf(self): return self._hsf @hsf.setter def hsf(self, now_radius): self._hsf = HaarSurroundFeature_new(now_radius) class HaarSurroundFeature_new: def __init__(self, r_inner, r_outer=None, val=None): if r_outer is None: r_outer = r_inner * 3 # print(r_outer) r_inner2 = r_inner * r_inner count_inner = r_inner2 count_outer = r_outer * r_outer - r_inner2 if val is None: val_inner = 1.0 / r_inner2 val_outer = -val_inner * count_inner / count_outer else: val_inner = val[0] val_outer = val[1] self.val_in = float(val_inner)#np.array(val_inner, dtype=np.float64) self.val_out = float(val_outer)#np.array(val_outer, dtype=np.float64) self.r_in = r_inner self.r_out = r_outer def get_kernel(self): # Defined here, but not yet used? # Create a kernel filled with the value of self.val_out kernel = np.ones(shape=(2 * self.r_out - 1, 2 * self.r_out - 1), dtype=np.float64) * self.val_out # Set the values of the inner area of the kernel using array slicing start = (self.r_out - self.r_in) end = (self.r_out + self.r_in - 1) kernel[start:end, start:end] = self.val_in return kernel @lru_cache(maxsize=lru_maxsize_vvs) def get_hsf_empty_array(len_sx,len_sy, frameint_x, frame_int_dtype, fcshape): # def get_hsf_empty_array(len_syx, frameint_x, frame_int_dtype, fcshape): # Function to reduce array allocation by providing an empty array first and recycling it with lru len_syx=(len_sy,len_sx) inner_sum = np.empty(len_syx, dtype=frame_int_dtype) # in_p_temp = np.empty((len_syx[0], frameint_x), dtype=frame_int_dtype) # in_p00 = np.empty(len_syx, dtype=frame_int_dtype) # in_p11 = np.empty(len_syx, dtype=frame_int_dtype) # in_p01 = np.empty(len_syx, dtype=frame_int_dtype) # in_p10 = np.empty(len_syx, dtype=frame_int_dtype) # inner_sum_temp = np.empty((*len_syx,4), dtype=frame_int_dtype) outer_sum = np.empty(len_syx, dtype=frame_int_dtype) # outer_sum_temp = np.empty((*len_syx,5), dtype=frame_int_dtype) p_temp = np.empty((len_sy, frameint_x), dtype=frame_int_dtype) p00 = np.empty(len_syx, dtype=frame_int_dtype) p11 = np.empty(len_syx, dtype=frame_int_dtype) p01 = np.empty(len_syx, dtype=frame_int_dtype) p10 = np.empty(len_syx, dtype=frame_int_dtype) response_list = np.empty(len_syx, dtype=np.float64)# or np.int32 frame_conv = np.zeros(shape=fcshape[0], dtype=np.uint8)# or np.float64 frame_conv_stride = frame_conv[::fcshape[1], ::fcshape[2]] return inner_sum, outer_sum, p_temp, p00, p11, p01, p10, response_list, frame_conv, frame_conv_stride # return inner_sum,in_p_temp,in_p00,in_p11,in_p01,in_p10, outer_sum, p_temp, p00, p11, p01, p10, response_list, frame_conv, frame_conv_stride @lru_cache(maxsize=lru_maxsize_vvs) def get_hsf_inout_index(padding, x_step, y_step, col, row, r_in, r_out):#,val_in,val_out): # y_steps,x_steps=np.ogrid[padding:y_step * len_sy + padding:y_step, padding:x_step * len_sx + padding:x_step] y_steps_arr = np.arange(padding, row - padding, y_step,dtype=np.int16) x_steps_arr = np.arange(padding, col - padding, x_step,dtype=np.int16) len_sx, len_sy = len(x_steps_arr), len(y_steps_arr) # y_steps_arr = np.arange(padding, row - padding, y_step) # x_steps_arr = np.arange(padding, col - padding, x_step) # len_sx, len_sy = len(x_steps_arr), len(y_steps_arr) # inarr_m = frame_int[y_steps[0]-r_in:y_steps[-1]-r_in+1:y_step] # inarr_p = frame_int[y_steps[0]+r_in:y_steps[-1]+r_in+1:y_step] # inarr_mm = inarr_m[:, x_steps[0]-r_in:x_steps[-1]-r_in+1:x_step] # inarr_mp = inarr_m[:, x_steps[0]+r_in:x_steps[-1]+r_in+1:x_step] # inarr_pm = inarr_p[:, x_steps[0]-r_in:x_steps[-1]-r_in+1:x_step] # inarr_pp = inarr_p[:, x_steps[0]+r_in:x_steps[-1]+r_in+1:x_step] # inarr_ym = np.arange(y_steps_arr[0]-r_in,y_steps_arr[-1]-r_in+1,y_step).reshape(-1,1)#np.ogrid[padding-r_in:y_steps[-1]-r_in+1:y_step],frame_int[y_steps[0]-r_in:y_steps[-1]-r_in+1:y_step] # inarr_yp = np.arange(y_steps_arr[0]+r_in,y_steps_arr[-1]+r_in+1,y_step).reshape(-1,1)#np.ogrid[padding+r_in:y_steps[-1]+r_in+1:y_step],frame_int[y_steps[0]+r_in:y_steps[-1]+r_in+1:y_step] # inarr_mm_index = (inarr_ym,np.arange(x_steps_arr[0]-r_in,x_steps_arr[-1]-r_in+1,x_step).reshape(1,-1))#inarr_m[:, x_steps[0]-r_in:x_steps[-1]-r_in+1:x_step] # inarr_mp_index = (inarr_ym,np.arange(x_steps_arr[0]+r_in,x_steps_arr[-1]+r_in+1,x_step).reshape(1,-1))#inarr_m[:, x_steps[0]+r_in:x_steps[-1]+r_in+1:x_step] # inarr_pm_index = (inarr_yp,inarr_mm_index[1].copy())#inarr_p[:, x_steps[0]-r_in:x_steps[-1]-r_in+1:x_step] # inarr_pp_index = (inarr_yp,inarr_mp_index[1].copy())#inarr_p[:, x_steps[0]+r_in:x_steps[-1]+r_in+1:x_step] # # y_ro_m = y_steps_arr - r_out # x_ro_m = x_steps_arr - r_out # y_ro_p = y_steps_arr + r_out # x_ro_p = x_steps_arr + r_out # # return x_steps_arr,y_steps_arr,len_sx,len_sy,inarr_mm_index,inarr_mp_index,inarr_pm_index,inarr_pp_index,y_ro_m,x_ro_m,y_ro_p,x_ro_p # y_rin_m_f = y_steps_arr[0] - r_in # y_rin_m_e = y_steps_arr[-1] - r_in + 1 # y_rin_p_f = y_steps_arr[0] + r_in # y_rin_p_e = y_steps_arr[-1] + r_in + 1 # # x_rin_m_f = x_steps_arr[0] - r_in # x_rin_m_e = x_steps_arr[-1] - r_in + 1 # x_rin_p_f = x_steps_arr[0] + r_in # x_rin_p_e = x_steps_arr[-1] + r_in + 1 # y_rin_m_f = padding - r_in # y_rin_m_e = y_steps_arr[-1] - r_in + 1 # y_rin_p_f = padding + r_in # y_rin_p_e = y_steps_arr[-1] + r_in + 1 # # x_rin_m_f = padding - r_in # x_rin_m_e = x_steps_arr[-1] - r_in + 1 # x_rin_p_f = padding + r_in # x_rin_p_e = x_steps_arr[-1] + r_in + 1 y_end=padding+(y_step*(len_sy-1)) x_end=padding+(x_step*(len_sx-1)) y_rin_m_f = padding - r_in y_rin_m_e = y_end - r_in + 1 y_rin_p_f = padding + r_in y_rin_p_e = y_end + r_in + 1 x_rin_m_f = padding - r_in x_rin_m_e = x_end - r_in + 1 x_rin_p_f = padding + r_in x_rin_p_e = x_end + r_in + 1 y_rin_m=slice(y_rin_m_f,y_rin_m_e,y_step) y_rin_p=slice(y_rin_p_f,y_rin_p_e,y_step) x_rin_m=slice(x_rin_m_f,x_rin_m_e,x_step) x_rin_p=slice(x_rin_p_f,x_rin_p_e,x_step) # y_rin_m=np.arange(padding-r_in,y_end-r_in+1,y_step,dtype=np.int16) # y_rin_p=np.arange(padding+r_in,y_end+r_in+1,y_step,dtype=np.int16) # x_rin_m=np.arange(padding-r_in,x_end-r_in+1,x_step,dtype=np.int16) # x_rin_p=np.arange(padding+r_in,x_end+r_in+1,x_step,dtype=np.int16) # y_ro_m = y_steps_arr - r_out # x_ro_m = x_steps_arr - r_out # y_ro_p = y_steps_arr + r_out # x_ro_p = x_steps_arr + r_out # y_ro_m = slice(max(0,y_steps_arr[0]-r_out),max(0,y_steps_arr[-1]-r_out),y_step)#,y_steps_arr - r_out # x_ro_m = slice(max(0,x_steps_arr[0]-r_out),max(0,x_steps_arr[-1]-r_out),x_step)#x_steps_arr - r_out # y_ro_p = slice(min(row,y_steps_arr[0]+r_out),min(row,y_steps_arr[-1]+r_out),y_step)#y_steps_arr + r_out # x_ro_p = slice(min(col,x_steps_arr[0]+r_out),min(col,x_steps_arr[-1]+r_out),x_step)#x_steps_arr + r_out # y_ro_m = np.clip(y_steps_arr - r_out,0,y_steps_arr[-1])#[:,np.newaxis] # x_ro_m = np.clip(x_steps_arr - r_out,0,x_steps_arr[-1])#[np.newaxis,:] # y_ro_p = np.clip(y_steps_arr + r_out,0,row)#[:,np.newaxis] # x_ro_p = np.clip(x_steps_arr + r_out,0,col)#[np.newaxis,:] y_ro_m = np.maximum(y_steps_arr - r_out,0)#[:,np.newaxis] x_ro_m = np.maximum(x_steps_arr - r_out,0)#[np.newaxis,:] y_ro_p = np.minimum(row,y_steps_arr + r_out)#[:,np.newaxis] x_ro_p = np.minimum(col,x_steps_arr + r_out)#[np.newaxis,:] # return x_steps_arr, y_steps_arr, len_sx, len_sy, y_rin_m_f, y_rin_m_e, y_rin_p_f, y_rin_p_e, x_rin_m_f, x_rin_m_e, x_rin_p_f, x_rin_p_e, y_ro_m, x_ro_m, y_ro_p, x_ro_p # return len_sx, len_sy, y_rin_m, y_rin_p, x_rin_m, x_rin_p, y_ro_m, x_ro_m, y_ro_p, x_ro_p,val_in,val_out,(row - 2 * padding, col - 2 * padding) return len_sx, len_sy, y_rin_m, y_rin_p, x_rin_m, x_rin_p, y_ro_m, x_ro_m, y_ro_p, x_ro_p,(row - 2 * padding, col - 2 * padding) @lru_cache(maxsize=lru_maxsize_s) def get_hsf_center(padding, x_step, y_step, min_loc):#min_x,min_y): # y_steps,x_steps=np.ogrid[padding:y_step * len_sy + padding:y_step, padding:x_step * len_sx + padding:x_step] # y_steps_arr = np.arange(padding, row - padding, y_step) # x_steps_arr = np.arange(padding, col - padding, x_step) # return x_steps_arr[min_x] - padding, y_steps_arr[min_y] - padding # return np.array(padding+(x_step*min_loc[0])-padding),np.array(padding+(y_step*min_loc[1])-padding) return padding+(x_step*min_loc[0])-padding,padding+(y_step*min_loc[1])-padding @lru_cache(maxsize=lru_maxsize_vvs) def get_frameint_empty_array(frame_shape,pad,x_step, y_step, r_in, r_out): frame_int_dtype=np.intc frame_pad=np.empty((frame_shape[0]+(pad*2),frame_shape[1]+(pad*2)),dtype=np.uint8) row,col=frame_pad.shape frame_int=np.empty((row+1,col+1),dtype=frame_int_dtype) y_steps_arr = np.arange(pad, row - pad, y_step,dtype=np.int16) x_steps_arr = np.arange(pad, col - pad, x_step,dtype=np.int16) len_sx, len_sy = len(x_steps_arr), len(y_steps_arr) y_end = pad + (y_step * (len_sy - 1)) x_end = pad + (x_step * (len_sx - 1)) y_rin_m = slice( pad - r_in, y_end - r_in + 1, y_step) y_rin_p = slice(pad + r_in, y_end + r_in + 1, y_step) x_rin_m = slice(pad - r_in, x_end - r_in + 1, x_step) x_rin_p = slice(pad + r_in, x_end + r_in + 1, x_step) in_p00_view=frame_int[y_rin_m,x_rin_m] in_p11_view=frame_int[y_rin_p,x_rin_p] in_p01_view=frame_int[y_rin_m,x_rin_p] in_p10_view=frame_int[y_rin_p,x_rin_m] y_ro_m = np.maximum(y_steps_arr - r_out,0)#[:,np.newaxis] x_ro_m = np.maximum(x_steps_arr - r_out,0)#[np.newaxis,:] y_ro_p = np.minimum(row,y_steps_arr + r_out)#[:,np.newaxis] x_ro_p = np.minimum(col,x_steps_arr + r_out)#[np.newaxis,:] return frame_pad,frame_int,in_p00_view,in_p11_view,in_p01_view,in_p10_view,y_ro_m, x_ro_m, y_ro_p, x_ro_p,(row - 2 * pad, col - 2 * pad),len_sx, len_sy # todo: Check performance when changing integer type numpy array to low bits integer type # todo: Consider using np.clip if the clamp function input meets some conditions # @profile def conv_int(frame_int, kernel, x_step,y_step, padding,in_p00_view, in_p11_view, in_p01_view, in_p10_view, y_ro_m, x_ro_m, y_ro_p, x_ro_p, f_shape, len_sx, len_sy): # , x_steps,y_steps):#xy_steps_list): """ :param frame_int: :param kernel: hsf :param step: (x,y) :param padding: int :return: """ # row, col = frame_int.shape # row -= 1 # col -= 1 # x_step, y_step = xy_step # padding2 = 2 * padding # f_shape = row - 2 * padding, col - 2 * padding # r_in = kernel.r_in # len_sx, len_sy = len(x_steps), len(y_steps) # inout_sum, p_temp, p_list, response_list, frameconvlist = get_hsf_empty_array((len_sy, len_sx), col + 1, # frame_int.dtype, (f_shape, y_step, x_step)) # inner_sum, outer_sum = inout_sum # p00, p11, p01, p10 = p_list # frame_conv, frame_conv_stride = frameconvlist # len_sx, len_sy, y_rin_m, y_rin_p, x_rin_m, x_rin_p, y_ro_m, x_ro_m, y_ro_p, x_ro_p,val_in,val_out,f_shape = get_hsf_inout_index( # padding, x_step, y_step, col, row, kernel.r_in, kernel.r_out,kernel.val_in,kernel.val_out) # len_sx, len_sy, y_rin_m, y_rin_p, x_rin_m, x_rin_p, y_ro_m, x_ro_m, y_ro_p, x_ro_p, f_shape = get_hsf_inout_index( # padding, x_step, y_step, col, row, kernel.r_in, kernel.r_out) inner_sum, outer_sum,p_temp, p00, p11, p01, p10, response_list, frame_conv, frame_conv_stride = get_hsf_empty_array(len_sx,len_sy,#(len_sy, len_sx), frame_int.shape[1],#col + 1, frame_int.dtype, ( f_shape, y_step, x_step)) # # inner_sum, in_p_temp, in_p00, in_p11, in_p01, in_p10, outer_sum, p_temp, p00, p11, p01, p10, response_list, frame_conv, frame_conv_stride= get_hsf_empty_array((len_sy, len_sx), # col + 1, # frame_int.dtype, ( # f_shape, y_step, # x_step)) # inout_sum, p_temp, p_list, response_list, frameconvlist = hsf_empty_array # inner_sum, outer_sum = inout_sum # p00, p11, p01, p10 = p_list # frame_conv, frame_conv_stride = frameconvlist # x_steps_st_end=np.asarray([x_steps[0],x_steps[len_sx-1]+1]) # # x_steps_st_end[1]+=1 # y_steps_st_end=np.asarray([y_steps[0],y_steps[len_sy-1]+1]) # y_steps_st_end[1] += 1 # x_steps_st_end=np.asarray([x_steps[0],x_steps[len_sx-1]+1]) # y_steps_st_end=np.asarray([y_steps[0],y_steps[len_sy-1]+1]) # xy_steps_st_end=np.asarray([[x_steps[0],x_steps[-1]],[y_steps[0],y_steps[-1]]]) # xy_steps_st_end[:,1]+=1 # xy_rin_m = xy_steps_st_end-r_in # xy_rin_p = xy_steps_st_end + r_in # xx==(y,x),m==MINUS,p==PLUS, ex: mm==(y-,x-) # inarr_m = frame_int[y_steps[0]-r_in:y_steps[-1]-r_in+1:y_step] # inarr_p = frame_int[y_steps[0]+r_in:y_steps[-1]+r_in+1:y_step] # inarr_mm = inarr_m[:, x_steps[0]-r_in:x_steps[-1]-r_in+1:x_step] # inarr_mp = inarr_m[:, x_steps[0]+r_in:x_steps[-1]+r_in+1:x_step] # inarr_pm = inarr_p[:, x_steps[0]-r_in:x_steps[-1]-r_in+1:x_step] # inarr_pp = inarr_p[:, x_steps[0]+r_in:x_steps[-1]+r_in+1:x_step] # y_rin_m = y_steps_st_end - r_in # x_rin_m = x_steps_st_end - r_in # y_rin_p = y_steps_st_end + r_in # x_rin_p = x_steps_st_end + r_in # # xx==(y,x),m==MINUS,p==PLUS, ex: mm==(y-,x-) # inarr_mm = frame_int[y_rin_m[0]:y_rin_m[1]:y_step, x_rin_m[0]:x_rin_m[1]:x_step] # inarr_mp = frame_int[y_rin_m[0]:y_rin_m[1]:y_step, x_rin_p[0]:x_rin_p[1]:x_step] # inarr_pm = frame_int[y_rin_p[0]:y_rin_p[1]:y_step, x_rin_m[0]:x_rin_m[1]:x_step] # inarr_pp = frame_int[y_rin_p[0]:y_rin_p[1]:y_step, x_rin_p[0]:x_rin_p[1]:x_step] # == inarr_mm + inarr_pp - inarr_mp - inarr_pm # inner_sum[:, :] = inarr_mm # inner_sum += inarr_pp # inner_sum -= inarr_mp # inner_sum -= inarr_pm # cv2.subtract(cv2.subtract(cv2.add(inarr_mm,inarr_pp),inarr_mp),inarr_pm,dst=inner_sum[:,:]) # inner_sum[:, :]=inarr_mm.__add__(inarr_pp).__sub__(inarr_mp).__sub__(inarr_pm) # inner_sum[:,:]=inarr_m[:, x_steps[0]-r_in:x_steps[-1]-r_in+1:x_step] # inner_sum += inarr_p[:, x_steps[0]+r_in:x_steps[-1]+r_in+1:x_step] # inner_sum -= inarr_m[:, x_steps[0]+r_in:x_steps[-1]+r_in+1:x_step] # inner_sum -= inarr_p[:, x_steps[0]-r_in:x_steps[-1]-r_in+1:x_step] # y_m_f=y_steps[0] - r_in # y_m_e=y_steps[-1] - r_in + 1 # y_p_f=y_steps[0] + r_in # y_p_e=y_steps[-1] + r_in + 1 # # x_m_f=x_steps[0] - r_in # x_m_e=x_steps[-1] - r_in + 1 # x_p_f=x_steps[0] + r_in # x_p_e=x_steps[-1] + r_in + 1 # inner_sum[:,:]=frame_int[y_steps[0]-r_in:y_steps[-1]-r_in+1:y_step, x_steps[0]-r_in:x_steps[-1]-r_in+1:x_step] # inner_sum += frame_int[y_steps[0]+r_in:y_steps[-1]+r_in+1:y_step, x_steps[0]+r_in:x_steps[-1]+r_in+1:x_step] # inner_sum -= frame_int[y_steps[0]-r_in:y_steps[-1]-r_in+1:y_step, x_steps[0]+r_in:x_steps[-1]+r_in+1:x_step] # inner_sum -= frame_int[y_steps[0]+r_in:y_steps[-1]+r_in+1:y_step, x_steps[0]-r_in:x_steps[-1]-r_in+1:x_step] # inner_sum[:, :] = frame_int[y_rin_m_f:y_rin_m_e:y_step, x_rin_m_f:x_rin_m_e:x_step] # inner_sum += frame_int[y_rin_p_f:y_rin_p_e:y_step, x_rin_p_f:x_rin_p_e:x_step] # inner_sum -= frame_int[y_rin_m_f:y_rin_m_e:y_step, x_rin_p_f:x_rin_p_e:x_step] # inner_sum -= frame_int[y_rin_p_f:y_rin_p_e:y_step, x_rin_m_f:x_rin_m_e:x_step] # inner_sum_temp[:, :,0] = frame_int[y_rin_m, x_rin_m].copy() # inner_sum_temp[:, :,1] = frame_int[y_rin_p, x_rin_p].copy() # inner_sum_temp[:, :,2] = -frame_int[y_rin_m, x_rin_p].copy() # inner_sum_temp[:, :,3] = -frame_int[y_rin_p, x_rin_m].copy() # cv2.transform(inner_sum_temp, np.ones((1, 4)), # dst=inner_sum) # # inner_sum[:, :] = frame_int[y_rin_m, x_rin_m] # inner_sum += frame_int[y_rin_p, x_rin_p] # inner_sum -= frame_int[y_rin_m, x_rin_p] # inner_sum -= frame_int[y_rin_p, x_rin_m] # inner_sum[:, :] = frame_int[y_rin_m, x_rin_m]+ frame_int[y_rin_p, x_rin_p]-frame_int[y_rin_m, x_rin_p]- frame_int[y_rin_p, x_rin_m] # inner_sum[:, :] = frame_int[y_rin_m, x_rin_m] + frame_int[y_rin_p, x_rin_p] - frame_int[y_rin_m, x_rin_p] - frame_int[y_rin_p, x_rin_m] inner_sum[:, :] = in_p00_view + in_p11_view - in_p01_view - in_p10_view # inarr_m = frame_int[inarr_ym] # inarr_p = frame_int[inarr_yp] # inner_sum[:, :] = inarr_m[:,inarr_xm]#inarr_mm # inner_sum += inarr_p[:,inarr_xp]#inarr_pp # inner_sum -= inarr_m[:,inarr_xp]#inarr_mp # inner_sum -= inarr_p[:,inarr_xm]#inarr_pm # Bottleneck here, I want to make it smarter. Someone do it. # (y,x) # p00=max(y_ro_m,0),max(x_ro_m,0) # p11=min(y_ro_p,ylim),min(x_ro_p,xlim) # p01=max(y_ro_m,0),min(x_ro_p,xlim) # p10=min(y_ro_p,ylim),max(x_ro_m,0) # y_ro_m = y_steps - kernel.r_out # x_ro_m = x_steps - kernel.r_out # y_ro_p = y_steps + kernel.r_out # x_ro_p = x_steps + kernel.r_out # p00 calc # np.take(frame_int, y_steps - kernel.r_out, axis=0, mode="clip", out=p_temp) # np.take(frame_int, y_ro_m, axis=0, mode="clip", out=p_temp) # np.take(p_temp, x_ro_m, axis=1, mode="clip", out=p00) # # p01 calc # np.take(p_temp, x_ro_p, axis=1, mode="clip", out=p01) # # p11 calc # np.take(frame_int, y_ro_p, axis=0, mode="clip", out=p_temp) # np.take(p_temp, x_ro_p, axis=1, mode="clip", out=p11) # # p10 calc # np.take(p_temp, x_ro_m, axis=1, mode="clip", out=p10) # np.take(frame_int, y_ro_m, axis=0, mode="clip", out=p_temp) # np.take(p_temp, x_ro_m, axis=1, mode="clip", out=p00) # # p01 calc # np.take(p_temp, x_ro_p, axis=1, mode="clip", out=p01) # # p11 calc # np.take(frame_int, y_ro_p, axis=0, mode="clip", out=p_temp) # np.take(p_temp, x_ro_p, axis=1, mode="clip", out=p11) # # p10 calc # np.take(p_temp, x_ro_m, axis=1, mode="clip", out=p10) frame_int.take( y_ro_m, axis=0, mode="clip", out=p_temp) p_temp.take(x_ro_m, axis=1, mode="clip", out=p00) # p01 calc p_temp.take( x_ro_p, axis=1, mode="clip", out=p01) # p11 calc frame_int.take( y_ro_p, axis=0, mode="clip", out=p_temp) p_temp.take( x_ro_p, axis=1, mode="clip", out=p11) # p10 calc p_temp.take( x_ro_m, axis=1, mode="clip", out=p10) # p_temp[:,:]=frame_int[y_ro_m.reshape(-1),:]#.copy() # p00[:,:]=np.asarray(p_temp[:,x_ro_m.reshape(-1)])#.copy() # # p01 calc # p01[:,:]=-p_temp[:,x_ro_p.reshape(-1)]#.copy() # # p11 calc # p_temp[:,:]=frame_int[y_ro_p.reshape(-1),:] # p11[:,:]=np.asarray(p_temp[:,x_ro_p.reshape(-1)])#.copy() # # p10 calc # p10[:,:]=-p_temp[:,x_ro_m.reshape(-1)]#.copy() # the point is this # p00=np.take(np.take(frame_int, y_ro_m, axis=0, mode="clip"), x_ro_m, axis=1, mode="clip") # p11=np.take(np.take(frame_int, y_ro_p, axis=0, mode="clip"), x_ro_p, axis=1, mode="clip") # p01=np.take(np.take(frame_int, y_ro_m, axis=0, mode="clip"), x_ro_p, axis=1, mode="clip") # p10=np.take(np.take(frame_int, y_ro_p, axis=0, mode="clip"), x_ro_m, axis=1, mode="clip") outer_sum[:, :] = p00 + p11 - p01 - p10 - inner_sum # cv2.transform(np.asarray([p00, p11, -p01, -p10, -inner_sum]).transpose((1, 2, 0)), np.ones((1, 5)), # dst=outer_sum) # https://answers.opencv.org/question/3120/how-to-sum-a-3-channel-matrix-to-a-one-channel-matrix/ # cv2.transform(np.asarray([frame_int[y_ro_m,x_ro_m], frame_int[y_ro_m,x_ro_p], -frame_int[y_ro_p,x_ro_p], -frame_int[y_ro_p,x_ro_m], -inner_sum]).transpose((1, 2, 0)), np.ones((1, 5)), # dst=outer_sum) # https://answers.opencv.org/question/3120/how-to-sum-a-3-channel-matrix-to-a-one-channel-matrix/ # cv2.transform(np.asarray([ # frame_int[y_ro_m,x_ro_m], # frame_int[y_ro_p,x_ro_p], # -frame_int[y_ro_m,x_ro_p], # -frame_int[y_ro_p,x_ro_m], # -inner_sum]).transpose((1, 2, 0)), np.ones((1, 5)), # dst=outer_sum) # https://answers.opencv.org/question/3120/how-to-sum-a-3-channel-matrix-to-a-one-channel-matrix/ # outer_sum_temp[:,:,0]=p00#frame_int[y_ro_m, x_ro_m] # outer_sum_temp[:, :, 1] = p11#frame_int[y_ro_p, x_ro_p] # outer_sum_temp[:, :, 2] = p01#frame_int[y_ro_m, x_ro_p] # outer_sum_temp[:, :, 3] = p10#frame_int[y_ro_p, x_ro_m] # p_temp[:,:]=frame_int[y_ro_m.reshape(-1),:]#.copy() # outer_sum_temp[:,:,0]=np.asarray(p_temp[:,x_ro_m.reshape(-1)])#frame_int[y_ro_m, x_ro_m] # outer_sum_temp[:, :, 2] = -p_temp[:,x_ro_p.reshape(-1)]#frame_int[y_ro_m, x_ro_p] # p_temp[:,:]=frame_int[y_ro_p.reshape(-1),:] # outer_sum_temp[:, :, 1] = np.asarray(p_temp[:,x_ro_p.reshape(-1)])#frame_int[y_ro_p, x_ro_p] # outer_sum_temp[:, :, 3] = -p_temp[:,x_ro_m.reshape(-1)]#frame_int[y_ro_p, x_ro_m] # outer_sum_temp[:, :, 4] = -inner_sum # cv2.transform(outer_sum_temp, np.ones((1, 5)), # dst=outer_sum) # https://answers.opencv.org/question/3120/how-to-sum-a-3-channel-matrix-to-a-one-channel-matrix/ # np.multiply(kernel.val_in, inner_sum, dtype=np.float64, out=response_list) # response_list += kernel.val_out * outer_sum cv2.addWeighted(inner_sum, kernel.val_in, outer_sum,# or p00 + p11 - p01 - p10 - inner_sum kernel.val_out, 0.0, dtype=cv2.CV_64F,#or cv2.CV_32S dst=response_list) # min_response, max_val, min_loc, max_loc = cv2.minMaxLoc(response_list) min_response, _, min_loc, _ = cv2.minMaxLoc(response_list) # center = ((x_steps_arr[min_loc[0]] - padding), (y_steps_arr[min_loc[1]] - padding)) # center = get_hsf_center(padding,x_step,y_step,min_loc)#[0],min_loc[1]) frame_conv_stride[:, :] = response_list # or # frame_conv_stride[:, :] = response_list.astype(np.uint8) # return frame_conv, min_response, center return frame_conv, min_response, get_hsf_center(padding,x_step,y_step,min_loc) @lru_cache(lru_maxsize_vvs) def get_ransac_frame(frame_shape): return np.empty(frame_shape,dtype=np.uint8),np.empty(frame_shape,dtype=np.uint8)#np.float64) @lru_cache(lru_maxsize_s) def get_center_noclamp(center_xy,radius): center_x, center_y = center_xy upper_x = center_x + radius lower_x = center_x - radius upper_y = center_y + radius lower_y = center_y - radius return center_x,center_y,upper_x,lower_x,upper_y,lower_y @lru_cache(lru_maxsize_s) def get_hsf_center_uplow(center_x,center_y,radius): hsf_center_x, hsf_center_y = center_x, center_y # ransac_xy_offset = (hsf_center_x-20, hsf_center_y-20) upper_x = hsf_center_x + max(20, radius) lower_x = hsf_center_x - max(20, radius) upper_y = hsf_center_y + max(20, radius) lower_y = hsf_center_y - max(20, radius) ransac_xy_offset = (lower_x, lower_y) return upper_x,lower_x,upper_y,lower_y,ransac_xy_offset class HSRAC_cls(object): def __init__(self): # I'd like to take into account things like print, end_time - start_time processing time, etc., but it's too much trouble. # For measuring total processing time self.main_start_time = timeit.default_timer() self.rng = np.random.default_rng() if old_mode: self.cvparam = CvParameters_old(default_radius, default_step) else: # os.environ["OPENBLAS_NUM_THREADS"]="1" # https://github.com/numpy/numpy/issues/22928 self.cvparam = CvParameters_new(default_radius, default_step) self.cv_modeo = ["first_frame", "radius_adjust", "blink_adjust", "normal"] self.now_modeo = self.cv_modeo[0] self.auto_radius_calc = AutoRadiusCalc() self.blink_detector = BlinkDetector() self.center_q1 = BlinkDetector() self.cap = None self.timedict = {"to_gray": [], "int_img": [], "conv_int": [], "crop": [], "ransac": [], "total_cv": []} # ransac self.rng = np.random.default_rng() self.sfc = np.random.default_rng(np.random.SFC64()) # self.kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (3, 3)) # or self.kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3)) self.gauss_k = cv2.getGaussianKernel(5, 0) def open_video(self, video_path): # Temporary implementation to run cap = cv2.VideoCapture(video_path) if not cap.isOpened(): raise IOError("Error opening video stream or file") self.cap = cap return True def read_frame(self): # Temporary implementation to run if not self.cap.isOpened(): return False ret, frame = self.cap.read() if ret: # I have set it to grayscale (1ch) just in case, but if the frame is 1ch, this line can be commented out. self.current_image = frame # debug code self.current_image_gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) return True return False # @profile def single_run(self): # Temporary implementation to run ## default_radius = 14 if imshow_enable or save_video: ori_frame = self.current_image.copy()# debug code # cropbox=[] # debug code blink_bd = False # frame = self.current_image_gray if self.now_modeo == self.cv_modeo[1]: # adjustment of radius # debug print # if calc_print_enable: # temp_radius = self.auto_radius_calc.get_radius() # print('Now radius:', temp_radius) # self.cvparam.radius = temp_radius self.cvparam.radius = self.auto_radius_calc.get_radius() if self.auto_radius_calc.adj_comp_flag: self.now_modeo = self.cv_modeo[2] if not skip_blink_detect else self.cv_modeo[3] radius, pad, step, hsf = self.cvparam.get_rpsh() # For measuring processing time of image processing cv_start_time = timeit.default_timer() frame = self.current_image_gray gray_frame = frame self.timedict["to_gray"].append(timeit.default_timer() - cv_start_time) # Calculate the integral image of the frame int_start_time = timeit.default_timer() if old_mode: # BORDER_CONSTANT is faster than BORDER_REPLICATE There seems to be almost no negative impact when BORDER_CONSTANT is used. frame_pad = cv2.copyMakeBorder(gray_frame, pad, pad, pad, pad, cv2.BORDER_CONSTANT) frame_int = cv2.integral(frame_pad) else: frame_pad, frame_int, in_p00_view, in_p11_view, in_p01_view, in_p10_view, y_ro_m, x_ro_m, y_ro_p, x_ro_p, f_shape, len_sx, len_sy = get_frameint_empty_array(gray_frame.shape,pad,step[0],step[1],hsf.r_in,hsf.r_out) cv2.copyMakeBorder(gray_frame, pad, pad, pad, pad, cv2.BORDER_CONSTANT,dst=frame_pad) cv2.integral(frame_pad,sum=frame_int,sdepth=cv2.CV_32S) self.timedict["int_img"].append(timeit.default_timer() - int_start_time) # Convolve the feature with the integral image conv_int_start_time = timeit.default_timer() if old_mode: xy_step = frameint_get_xy_step_old(frame_int.shape, step, pad, start_offset=None, end_offset=None) frame_conv, response, center_xy = conv_int_old(frame_int, hsf, step, pad, xy_step) else: # frame_conv, response, center_xy = conv_int(frame_int, hsf, step, pad) # , x_step,y_step) frame_conv, response, center_xy = conv_int(frame_int, hsf, step[0],step[1], pad,in_p00_view, in_p11_view, in_p01_view, in_p10_view, y_ro_m, x_ro_m, y_ro_p, x_ro_p, f_shape, len_sx, len_sy) # , x_step,y_step) # x_step,y_step = frameint_get_xy_step(frame_int.shape, step, pad, start_offset=None, end_offset=None) self.timedict["conv_int"].append(timeit.default_timer() - conv_int_start_time) crop_start_time = timeit.default_timer() # Define the center point and radius # center_x, center_y = center_xy # upper_x = center_x + radius # lower_x = center_x - radius # upper_y = center_y + radius # lower_y = center_y - radius center_x, center_y, upper_x, lower_x, upper_y, lower_y=get_center_noclamp(center_xy,radius) # Crop the image using the calculated bounds cropped_image = safe_crop(gray_frame, lower_x, lower_y, upper_x, upper_y) # cropbox=[clamp(val, 0, gray_frame.shape[i]) for i,val in zip([1,0,1,0],[lower_x,lower_y,upper_x,upper_y])] # debug code if self.now_modeo == self.cv_modeo[0] or self.now_modeo == self.cv_modeo[1]: # If mode is first_frame or radius_adjust, record current radius and response self.auto_radius_calc.add_response(radius, response) elif self.now_modeo == self.cv_modeo[2]: # Statistics for blink detection if self.blink_detector.response_len() < blink_init_frames: self.blink_detector.add_response(cv2.mean(cropped_image)[0]) upper_x = center_x + max(20, radius) lower_x = center_x - max(20, radius) upper_y = center_y + max(20, radius) lower_y = center_y - max(20, radius) self.center_q1.add_response( cv2.mean(safe_crop(gray_frame, lower_x, lower_y, upper_x, upper_y, keepsize=False))[ 0 ] ) else: self.blink_detector.calc_thresh() self.center_q1.calc_thresh() self.now_modeo = self.cv_modeo[3] else: if 0 in cropped_image.shape: # This line may not be needed. The image will be cropped using safecrop. # If shape contains 0, it is not detected well. print("Something's wrong.") else: orig_x, orig_y = center_x, center_y if self.blink_detector.enable_detect_flg: # If the average value of cropped_image is greater than response_max # (i.e., if the cropimage is whitish if self.blink_detector.detect(cv2.mean(cropped_image)[0]): # blink print("BLINK BD") blink_bd = True # if imshow_enable or save_video: # cv2.circle(frame, (orig_x, orig_y), 6, (0, 0, 255), -1) # cv2.circle(ori_frame, (center_x, center_y), 7, (255, 0, 0), -1) # If you want to update response_max. it may be more cost-effective to rewrite response_list in the following way # https://stackoverflow.com/questions/42771110/fastest-way-to-left-cycle-a-numpy-array-like-pop-push-for-a-queue # cv_end_time = timeit.default_timer() self.timedict["crop"].append(timeit.default_timer() - crop_start_time) # self.timedict["total_cv"].append(cv_end_time - cv_start_time) # if calc_print_enable: # the lower the response the better the likelyhood of there being a pupil. you can adujst the radius and steps accordingly # print('Kernel response:', response) # print('Pixel position:', center_xy) # # if imshow_enable: # if self.now_modeo != self.cv_modeo[0] and self.now_modeo != self.cv_modeo[1]: # if 0 in cropped_image.shape: # If shape contains 0, it is not detected well. # pass # else: # cv2.imshow("crop", cropped_image) # cv2.imshow("frame", frame) # if cv2.waitKey(1) & 0xFF == ord("q"): # pass if self.now_modeo == self.cv_modeo[0]: # Moving from first_frame to the next mode if skip_autoradius and skip_blink_detect: self.now_modeo = self.cv_modeo[3] elif skip_autoradius: self.now_modeo = self.cv_modeo[2] else: self.now_modeo = self.cv_modeo[1] # For measuring processing time of image processing ransac_start_time = timeit.default_timer() # Crop first to reduce the amount of data to process. # frame = cropped_image[0:len(cropped_image) - 10, :] # To reduce the processing data, first convert to 1-channel and then blur. # The processing results were the same when I swapped the order of blurring and 1-channelization. # frame_gray = cv2.GaussianBlur(frame, (5, 5), 0) # cv2.GaussianBlur is slow (use 10%) # use cv2.blur() or cv2.boxFilter()? # or # frame_gray =cv2.boxFilter(frame, -1,(5, 5))# https://github.com/bfraboni/FastGaussianBlur # cv2.boxFilter(frame_gray, -1,(5, 5),dst=frame_gray) # cv2.boxFilter(frame_gray, -1,(5, 5),dst=frame_gray) if old_mode: frame_gray = cv2.GaussianBlur(frame, (5, 5), 0) else: frame_gray = cv2.sepFilter2D(frame, -1, self.gauss_k, self.gauss_k) #todo:no numpy and use lru # hsf_center_x, hsf_center_y = center_x.copy(), center_y.copy() # # ransac_xy_offset = (hsf_center_x-20, hsf_center_y-20) # upper_x = hsf_center_x + max(20, radius) # lower_x = hsf_center_x - max(20, radius) # upper_y = hsf_center_y + max(20, radius) # lower_y = hsf_center_y - max(20, radius) # ransac_xy_offset = (lower_x, lower_y) upper_x, lower_x, upper_y, lower_y, ransac_xy_offset = get_hsf_center_uplow(center_x,center_y,radius) # Crop the image using the calculated bounds #todo:safecrop tune frame_gray_crop = safe_crop(frame_gray, lower_x, lower_y, upper_x, upper_y) th_frame,fic_frame=get_ransac_frame(frame_gray_crop.shape) frame = frame_gray_crop # todo: It can cause bugs. # this will need to be adjusted everytime hardware is changed (brightness of IR, Camera postion, etc)m # min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(frame_gray_crop) min_val = cv2.minMaxLoc(frame_gray_crop)[0] # min_val=cv2.reduce(frame_gray_crop,1,cv2.REDUCE_MIN).min() # threshold_value = min_val + thresh_add if old_mode: _, thresh = cv2.threshold(frame_gray_crop, min_val + thresh_add, 255, cv2.THRESH_BINARY) else: cv2.threshold(frame_gray_crop, min_val + thresh_add, 255, cv2.THRESH_BINARY, dst=th_frame) # print(thresh.shape, frame_gray.shape) try: if old_mode: opening = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, self.kernel) closing = cv2.morphologyEx(opening, cv2.MORPH_CLOSE, self.kernel) th_frame = 255 - closing else: cv2.morphologyEx(th_frame, cv2.MORPH_OPEN, self.kernel,dst=fic_frame) cv2.morphologyEx(fic_frame, cv2.MORPH_CLOSE, self.kernel,dst=fic_frame) cv2.bitwise_not(fic_frame,fic_frame) # th_frame = 255 - closing except Exception as e: raise e # I want to eliminate try here because try tends to be slow in execution. fic_frame = 255 - frame_gray_crop if old_mode: contours, _ = cv2.findContours(th_frame, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE) else: contours = cv2.findContours(fic_frame, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)[0] # contours, _ = cv2.findContours(fic_frame, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE) # or # contours, _=cv2.findContours(th_frame, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE) # cv2.findContours(th_frame, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE) # cv2.findContours(th_frame, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE) # cv2.findContours(th_frame, cv2.RETR_LIST, cv2.CHAIN_APPROX_NONE) if not blink_bd and self.blink_detector.enable_detect_flg: threshold_value = self.center_q1.quartile_1 if threshold_value < min_val + thresh_add: # In most of these cases, the pupil is at the edge of the eye. if old_mode: thresh = cv2.threshold(frame_gray_crop, (min_val + thresh_add * 4 + threshold_value) / 2, 255, cv2.THRESH_BINARY)[1] else: cv2.threshold(frame_gray_crop, (min_val + thresh_add * 4 + threshold_value) / 2, 255, cv2.THRESH_BINARY,dst=th_frame) else: threshold_value = self.center_q1.quartile_1 if old_mode: _, thresh = cv2.threshold(frame_gray_crop, threshold_value, 255, cv2.THRESH_BINARY) else: cv2.threshold(frame_gray_crop, threshold_value, 255, cv2.THRESH_BINARY,dst=th_frame) try: if old_mode: opening = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, self.kernel) closing = cv2.morphologyEx(opening, cv2.MORPH_CLOSE, self.kernel) th_frame = 255 - closing else: cv2.morphologyEx(th_frame, cv2.MORPH_OPEN, self.kernel, dst=fic_frame) cv2.morphologyEx(fic_frame, cv2.MORPH_CLOSE, self.kernel, dst=fic_frame) cv2.bitwise_not(fic_frame, fic_frame) except Exception as e: raise e # I want to eliminate try here because try tends to be slow in execution. fic_frame = 255 - frame_gray_crop if old_mode: contours2, _ = cv2.findContours(th_frame, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE) contours = (*contours, *contours2) else: # contours2, _ = cv2.findContours(fic_frame, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)#cv2.CHAIN_APPROX_NONE) # contours2 = cv2.findContours(fic_frame, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)[0] contours = (*contours, *cv2.findContours(fic_frame, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)[0]) # contours = (*contours, *contours2) # hull = [] # # This way is faster than contours[i] # # But maybe this one is faster. hull = [cv2.convexHull(cnt, False) for cnt in contours] # for cnt in contours: # # pass # hull.append(cv2.convexHull(cnt, False)) if not contours: # If empty, go to next loop return int(center_x), int(center_y), th_frame, frame, gray_frame if old_mode: hull = [cv2.convexHull(cnt, False) for cnt in contours] else: cnt_ind = None max_area = -1 for i, cnt in enumerate(contours): now_area = cv2.contourArea(cnt) if max_area < now_area: max_area = now_area cnt_ind = i hull = cv2.convexHull(contours[cnt_ind], False) # if not hull: # If empty, go to next loop # return int(center_x), int(center_y), th_frame, frame, gray_frame if 1: if old_mode: cnt = sorted(hull, key=cv2.contourArea) maxcnt = cnt[-1] else: maxcnt = hull # ellipse = cv2.fitEllipse(maxcnt) if old_mode: ransac_data = fit_rotated_ellipse_ransac_old(maxcnt.reshape(-1, 2), self.rng) else: ransac_data = fit_rotated_ellipse_ransac(maxcnt.reshape(-1, 2).astype(np.float64), self.sfc) if ransac_data is None: # ransac_data is None==maxcnt.shape[0]= 2.1 * h: # new blink detection algo lmao this works pretty good actually print("RAN BLINK") # return center_x, center_y, frame, frame, True # cx = center_x - (csx - cx) # we find the difference between the crop size and ransac point, and subtract from the center point from HSF # cy = center_y - (csy - cy) # csy = frame.shape[0] # csx = frame.shape[1] csy = gray_frame.shape[0] csx = gray_frame.shape[1] # cx = clamp((cx - 20) + center_x, 0, csx) # cy = clamp((cy - 20) + center_y, 0, csy) cx = int(clamp(cx + ransac_xy_offset[0], 0, csx)) cy = int(clamp(cy + ransac_xy_offset[1], 0, csy)) # cv_end_time = timeit.default_timer() if imsave_flg: cv2.circle(ori_frame, (int(center_x), int(center_y)), 3, (0, 255, 0), -1) cv2.drawContours(ori_frame, contours, -1, (255, 0, 0), 1) cv2.circle(ori_frame, (int(cx), int(cy)), 2, (0, 0, 255), -1) # cx1, cy1, w1, h1, theta1 = fit_rotated_ellipse(maxcnt.reshape(-1, 2)) # cv2.ellipse( # ori_frame, # (cx, cy), # (int(w), int(h)), # theta * 180.0 / np.pi, # 0.0, # 360.0, # (50, 250, 200), # 1, # ) # cv2.imshow("crop", cropped_image) # cv2.imshow("frame", frame) if imshow_enable: cv2.imshow("ori_frame", ori_frame) if cv2.waitKey(1) & 0xFF == ord("q"): pass # except Exception as e: # print(e) # pass # debug code # try: # if any([isinstance(val, float) for val in [cx, cy]]): # print() # return int(cx), int(cy),cropbox, ori_frame,thresh, frame, gray_frame # except: # if any([isinstance(val, float) for val in [center_x, center_y]]): # print() # return center_x, center_y,cropbox, ori_frame,thresh, frame, gray_frame # print(frame_gray.shape, thresh.shape) cv_end_time = timeit.default_timer() self.timedict["ransac"].append(cv_end_time - ransac_start_time) self.timedict["total_cv"].append(cv_end_time - cv_start_time) try: return int(cx), int(cy), th_frame, frame, gray_frame except: return int(center_x), int(center_y), th_frame, frame, gray_frame if __name__ == "__main__": # print(np.show_config()) logger.info(this_file_basename) if save_logfile: logger.info("log path: {}".format(logfilename)) logger.info("alg ver: {}".format(alg_ver)) logger.info("alg mode: {}".format("old" if old_mode else "new")) logger.info("loops: {}".format(loop_num)) logger.info("video name: {}".format(os.path.basename(input_video_path))) cap = cv2.VideoCapture(input_video_path) logger.info("video info: size:{}x{} fps:{} frames:{} total:{:.3f} sec".format(int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)), int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)), cap.get(cv2.CAP_PROP_FPS), int(cap.get(cv2.CAP_PROP_FRAME_COUNT)), cap.get(cv2.CAP_PROP_FRAME_COUNT) / cap.get(cv2.CAP_PROP_FPS))) cap.release() if not print_enable: def print(*args, **kwargs): pass hsrac = HSRAC_cls() # For measuring total processing time main_start_time = timeit.default_timer() for i in range(loop_num): hsrac.open_video(input_video_path) while hsrac.read_frame(): if imsave_flg: base_gray = hsrac.current_image_gray.copy() base_img = hsrac.current_image.copy() cv2.imshow("frame", base_gray) hsf_x, hsf_y, hsf_cropbox, *_ = hsrac.single_run() # # hsrac_x, hsrac_y, hsrac_cropbox, *_ = er_hsracs.run(base_gray) # if 0:#random.random()<0.1: # hsrac_x, hsrac_y, hsrac_cropbox, *_ = er_hsracs.run(cv2.resize(base_gray,None,fx=0.75,fy=0.75).copy()) # hsrac_x=int(hsrac_x*1.25) # hsrac_y=int(hsrac_y*1.25) # hsrac_cropbox=[int(val*1.25) for val in hsrac_cropbox] # else: # hsrac_x, hsrac_y, hsrac_cropbox,ori_frame, *_ = er_hsracs.run(base_gray) # cv2.rectangle(base_img, hsf_cropbox[:2], hsf_cropbox[2:], (0, 0, 255), 3) # cv2.rectangle(base_img, hsrac_cropbox[:2], hsrac_cropbox[2:], (255, 0, 0), 1) cv2.circle(base_img, (hsf_x, hsf_y), 6, (0, 0, 255), -1) # try: # cv2.circle(base_img, (hsrac_x, hsrac_y), 3, (255, 0, 0), -1) # except: # print() if imshow_enable: cv2.imshow("hsf_hsrac", base_img) if save_video: video_wr.write(cv2.resize(base_img, (200, 150))) if cv2.waitKey(1) & 0xFF == ord("q"): pass else: _ = hsrac.single_run() if save_video: video_wr.release() hsrac.cap.release() cv2.destroyAllWindows() main_end_time = timeit.default_timer() main_total_time = main_end_time - main_start_time if not print_enable: # del print # or print = __builtins__.print logger.info("") for k, v in hsrac.timedict.items(): # number=1, precision=5 len_v = len(v) best = min(v) # / number worst = max(v) # / number logger.info(k + ":") logger.info(TimeitResult(loop_num, len_v, best, worst, v, 5)) logger.info(FPSResult(loop_num, len_v, worst, best, v, 5)) # print("") logger.info("") logger.info(f"{this_file_basename}: ALL Finish {format_time(main_total_time)}")