Add support for TensorFlow Lite for Microcontrollers

This has been a long-time coming, but, it's finally here. TensorFlow lite
runs on the OpenMV Cam now. Better yet, a person detection model
is now built-into all OpenMV Cams too!

Our default code does image classification and supports multi-scale
object detection using a sliding window.

In a coming PR I will add mobilenet examples for the H7 with SDRAM
and image segmentation support.
This commit is contained in:
Kwabena W. Agyeman 2019-10-23 00:20:54 -07:00 committed by iabdalkader
parent 1915b9142f
commit 6a9c0a370e
27 changed files with 1182 additions and 6 deletions

View File

@ -0,0 +1,49 @@
# TensorFlow Lite Person Dection Example
#
# Google's Person Detection Model detects if a person is in view.
#
# In this example we slide the detector window over the image and get a list
# of activations. Note that use a CNN with a sliding window is extremely compute
# expensive so for an exhaustive search do not expect the CNN to be real-time.
import sensor, image, time, os, tf
sensor.reset() # Reset and initialize the sensor.
sensor.set_pixformat(sensor.GRAYSCALE) # Set pixel format to RGB565 (or GRAYSCALE)
sensor.set_framesize(sensor.QVGA) # Set frame size to QVGA (320x240)
sensor.set_windowing((240, 240)) # Set 240x240 window.
sensor.skip_frames(time=2000) # Let the camera adjust.
# Load the built-in person detection network (the network is in your OpenMV Cam's firmware).
net = tf.load('person_detection')
labels = ['unsure', 'person', 'no_person']
clock = time.clock()
while(True):
clock.tick()
img = sensor.snapshot()
# net.classify() will run the network on an roi in the image (or on the whole image if the roi is not
# specified). A classification score output vector will be generated for each location. At each scale the
# detection window is moved around in the ROI using x_overlap (0-1) and y_overlap (0-1) as a guide.
# If you set the overlap to 0.5 then each detection window will overlap the previous one by 50%. Note
# the computational work load goes WAY up the more overlap. Finally, for multi-scale matching after
# sliding the network around in the x/y dimensions the detection window will shrink by scale_mul (0-1)
# down to min_scale (0-1). For example, if scale_mul is 0.5 the detection window will shrink by 50%.
# Note that at a lower scale there's even more area to search if x_overlap and y_overlap are small...
# Setting x_overlap=-1 forces the window to stay centered in the ROI in the x direction always. If
# y_overlap is not -1 the method will search in all vertical positions.
# Setting y_overlap=-1 forces the window to stay centered in the ROI in the y direction always. If
# x_overlap is not -1 the method will serach in all horizontal positions.
# default settings just do one detection... change them to search the image...
for obj in net.classify(img, min_scale=0.5, scale_mul=0.5, x_overlap=-1, y_overlap=-1):
print("**********\nDetections at [x=%d,y=%d,w=%d,h=%d]" % obj.rect())
for i in range(len(obj.output())):
print("%s = %f" % (labels[i], obj.output()[i]))
img.draw_rectangle(obj.rect())
img.draw_string(obj.x()+3, obj.y()-1, labels[obj.output().index(max(obj.output()))], mono_space = False)
print(clock.fps(), "fps")

View File

@ -0,0 +1,43 @@
# TensorFlow Lite Person Dection Example
#
# Google's Person Detection Model detects if a person is in view.
#
# In this example we slide the detector window over the image and get a list
# of activations. Note that use a CNN with a sliding window is extremely compute
# expensive so for an exhaustive search do not expect the CNN to be real-time.
import sensor, image, time, os, tf
sensor.reset() # Reset and initialize the sensor.
sensor.set_pixformat(sensor.GRAYSCALE) # Set pixel format to RGB565 (or GRAYSCALE)
sensor.set_framesize(sensor.QVGA) # Set frame size to QVGA (320x240)
sensor.set_windowing((240, 240)) # Set 240x240 window.
sensor.skip_frames(time=2000) # Let the camera adjust.
# Load the built-in person detection network (the network is in your OpenMV Cam's firmware).
net = tf.load('person_detection')
labels = ['unsure', 'person', 'no_person']
clock = time.clock()
while(True):
clock.tick()
img = sensor.snapshot()
# net.classify() will run the network on an roi in the image (or on the whole image if the roi is not
# specified). A classification score output vector will be generated for each location. At each scale the
# detection window is moved around in the ROI using x_overlap (0-1) and y_overlap (0-1) as a guide.
# If you set the overlap to 0.5 then each detection window will overlap the previous one by 50%. Note
# the computational work load goes WAY up the more overlap. Finally, for multi-scale matching after
# sliding the network around in the x/y dimensions the detection window will shrink by scale_mul (0-1)
# down to min_scale (0-1). For example, if scale_mul is 0.5 the detection window will shrink by 50%.
# Note that at a lower scale there's even more area to search if x_overlap and y_overlap are small...
# default settings just do one detection... change them to search the image...
for obj in net.classify(img, min_scale=1.0, scale_mul=0.5, x_overlap=0.0, y_overlap=0.0):
print("**********\nDetections at [x=%d,y=%d,w=%d,h=%d]" % obj.rect())
for i in range(len(obj.output())):
print("%s = %f" % (labels[i], obj.output()[i]))
img.draw_rectangle(obj.rect())
img.draw_string(obj.x()+3, obj.y()-1, labels[obj.output().index(max(obj.output()))], mono_space = False)
print(clock.fps(), "fps")

View File

@ -16,6 +16,7 @@ endif
# Commands
CC = $(Q)arm-none-eabi-gcc
CXX = $(Q)arm-none-eabi-g++
AS = $(Q)arm-none-eabi-as
LD = $(Q)arm-none-eabi-ld
AR = $(Q)arm-none-eabi-ar
@ -50,6 +51,7 @@ MICROPY_DIR=micropython
OMV_DIR=omv
LEPTON_DIR=lepton
MLX_DIR=mlx
TENSORFLOW_DIR=libtf
WINC1500_DIR=winc1500
BOOTLDR_DIR=bootloader
UVC_DIR=uvc
@ -119,6 +121,7 @@ OMV_CFLAGS += -I$(TOP_DIR)/$(OMV_DIR)/img/
OMV_CFLAGS += -I$(OMV_BOARD_CONFIG_DIR)
OMV_CFLAGS += -I$(TOP_DIR)/$(LEPTON_DIR)/include/
OMV_CFLAGS += -I$(TOP_DIR)/$(MLX_DIR)/include/
OMV_CFLAGS += -I$(TOP_DIR)/$(TENSORFLOW_DIR)/$(CPU)/
OMV_CFLAGS += -I$(TOP_DIR)/$(WINC1500_DIR)/include/
UVC_CFLAGS = $(CFLAGS)
@ -261,6 +264,7 @@ FIRM_OBJ += $(addprefix $(BUILD)/$(OMV_DIR)/py/, \
py_winc.o \
py_cpufreq.o \
py_nn.o \
py_tf.o \
)
@ -448,6 +452,8 @@ ifeq ($(CUBEAI), 1)
include $(TOP_DIR)/stm32cubeai/cube.mk
endif
FIRM_OBJ += $(wildcard $(TOP_DIR)/$(TENSORFLOW_DIR)/$(CPU)/*.a)
# Bootloader object files
BOOT_OBJ += $(wildcard $(BUILD)/$(BOOTLDR_DIR)/src/*.o)
BOOT_OBJ += $(wildcard $(BUILD)/$(STHAL_DIR)/src/*.o)

203
src/libtf/cortex-m4/LICENSE Normal file
View File

@ -0,0 +1,203 @@
Copyright 2019 The TensorFlow Authors. All rights reserved.
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

View File

@ -0,0 +1,5 @@
You must link this library to your application with arm-none-eabi-g++ and have implemented puts().
C Compile Flags: -D __FPU_PRESENT=1 -DGEMMLOWP_ALLOW_SLOW_SCALAR_FALLBACK -DNDEBUG -DTF_LITE_DISABLE_X86_NEON -DTF_LITE_MCU_DEBUG_LOG -DTF_LITE_STATIC_MEMORY -MMD -O3 -Wall -Wextra -Wvla -Wno-missing-field-initializers -Wno-parentheses -Wno-sign-compare -Wno-strict-aliasing -Wno-type-limits -Wno-unused-but-set-variable -Wno-unused-parameter -Wno-unused-variable -Wno-write-strings -fdata-sections -ffunction-sections -fmessage-length=0 -fomit-frame-pointer -funsigned-char -fno-builtin -fno-delete-null-pointer-checks -fno-exceptions -fno-unwind-tables -mfloat-abi=hard -mlittle-endian -mthumb -mno-unaligned-access -nostdlib -Wno-pointer-sign -DARM_CMSIS_NN_M4 -DARM_MATH_CM4 -mcpu=cortex-m4 -mfpu=fpv4-sp-d16 -mtune=cortex-m4
CXX Compile Flags: -D __FPU_PRESENT=1 -DGEMMLOWP_ALLOW_SLOW_SCALAR_FALLBACK -DNDEBUG -DTF_LITE_DISABLE_X86_NEON -DTF_LITE_MCU_DEBUG_LOG -DTF_LITE_STATIC_MEMORY -MMD -O3 -Wall -Wextra -Wvla -Wno-missing-field-initializers -Wno-parentheses -Wno-sign-compare -Wno-strict-aliasing -Wno-type-limits -Wno-unused-but-set-variable -Wno-unused-parameter -Wno-unused-variable -Wno-write-strings -fdata-sections -ffunction-sections -fmessage-length=0 -fomit-frame-pointer -funsigned-char -fno-builtin -fno-delete-null-pointer-checks -fno-exceptions -fno-unwind-tables -mfloat-abi=hard -mlittle-endian -mthumb -mno-unaligned-access -nostdlib -std=c++11 -std=gnu++11 -fno-rtti -fpermissive -DARM_CMSIS_NN_M4 -DARM_MATH_CM4 -mcpu=cortex-m4 -mfpu=fpv4-sp-d16 -mtune=cortex-m4

BIN
src/libtf/cortex-m4/libtf.a Normal file

Binary file not shown.

View File

@ -0,0 +1,61 @@
/* This file is part of the OpenMV project.
* Copyright (c) 2013-2019 Ibrahim Abdelkader <iabdalkader@openmv.io> & Kwabena W. Agyeman <kwagyeman@openmv.io>
* This work is licensed under the MIT license, see the file LICENSE for details.
*/
#ifndef __LIBTF_H
#define __LIBTF_H
#ifdef __cplusplus
extern "C" {
#endif
// Call this first to get the shape of the model input.
// Returns 0 on success and 1 on failure.
// Errors are printed to stdout.
int libtf_get_input_data_hwc(const unsigned char *model_data, // TensorFlow Lite binary model (8-bit quant).
unsigned char *tensor_arena, // As big as you can make it scratch buffer.
const unsigned int tensor_arena_size, // Size of the above scratch buffer.
unsigned int *input_height, // Height for the model.
unsigned int *input_width, // Width for the model.
unsigned int *input_channels); // Channels for the model (1 for grayscale8 and 3 for rgb888).
// Call this second to get the shape of the model output.
// Returns 0 on success and 1 on failure.
// Errors are printed to stdout.
int libtf_get_output_data_hwc(const unsigned char *model_data, // TensorFlow Lite binary model (8-bit quant).
unsigned char *tensor_arena, // As big as you can make it scratch buffer.
const unsigned int tensor_arena_size, // Size of the above scratch buffer.
unsigned int *output_height, // Height for the model.
unsigned int *output_width, // Width for the model.
unsigned int *output_channels); // Channels for the model (1 for grayscale8 and 3 for rgb888).
// Callback to populate the model input data byte array (laid out in [height][width][channel] order).
typedef void (*libtf_input_data_callback_t)(void *callback_data,
unsigned char *model_input,
const unsigned int input_height,
const unsigned int input_width,
const unsigned int input_channels);
// Callback to use the model output data byte array (laid out in [height][width][channel] order).
typedef void (*libtf_output_data_callback_t)(void *callback_data,
unsigned char *model_output,
const unsigned int output_height,
const unsigned int output_width,
const unsigned int output_channels);
// Returns 0 on success and 1 on failure.
// Errors are printed to stdout.
int libtf_invoke(const unsigned char *model_data, // TensorFlow Lite binary model (8-bit quant).
unsigned char *tensor_arena, // As big as you can make it scratch buffer.
const unsigned int tensor_arena_size, // Size of the above scratch buffer.
libtf_input_data_callback_t input_callback, // Callback to populate the model input data byte array.
void *input_callback_data, // User data structure passed to input callback.
libtf_output_data_callback_t output_callback, // Callback to use the model output data byte array.
void *output_callback_data); // User data structure passed to output callback.
#ifdef __cplusplus
}
#endif
#endif // __LIBTF_H

Binary file not shown.

View File

@ -0,0 +1,27 @@
/* Copyright 2019 The TensorFlow Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
// This is a standard TensorFlow Lite model file that has been converted into a
// C data array, so it can be easily compiled into a binary for devices that
// don't have a file system. It was created using the command:
// xxd -i person_detect.tflite > person_detect_model_data.cc
#ifndef TENSORFLOW_LITE_EXPERIMENTAL_MICRO_EXAMPLES_PERSON_DETECTION_PERSON_DETECT_MODEL_DATA_H_
#define TENSORFLOW_LITE_EXPERIMENTAL_MICRO_EXAMPLES_PERSON_DETECTION_PERSON_DETECT_MODEL_DATA_H_
extern const unsigned char g_person_detect_model_data[];
extern const int g_person_detect_model_data_len;
#endif // TENSORFLOW_LITE_EXPERIMENTAL_MICRO_EXAMPLES_PERSON_DETECTION_PERSON_DETECT_MODEL_DATA_H_

203
src/libtf/cortex-m7/LICENSE Normal file
View File

@ -0,0 +1,203 @@
Copyright 2019 The TensorFlow Authors. All rights reserved.
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

View File

@ -0,0 +1,5 @@
You must link this library to your application with arm-none-eabi-g++ and have implemented puts().
C Compile Flags: -D __FPU_PRESENT=1 -DGEMMLOWP_ALLOW_SLOW_SCALAR_FALLBACK -DNDEBUG -DTF_LITE_DISABLE_X86_NEON -DTF_LITE_MCU_DEBUG_LOG -DTF_LITE_STATIC_MEMORY -MMD -O3 -Wall -Wextra -Wvla -Wno-missing-field-initializers -Wno-parentheses -Wno-sign-compare -Wno-strict-aliasing -Wno-type-limits -Wno-unused-but-set-variable -Wno-unused-parameter -Wno-unused-variable -Wno-write-strings -fdata-sections -ffunction-sections -fmessage-length=0 -fomit-frame-pointer -funsigned-char -fno-builtin -fno-delete-null-pointer-checks -fno-exceptions -fno-unwind-tables -mfloat-abi=hard -mlittle-endian -mthumb -mno-unaligned-access -nostdlib -Wno-pointer-sign -DARM_CMSIS_NN_M7 -DARM_MATH_CM7 -mcpu=cortex-m7 -mfpu=fpv5-sp-d16 -mtune=cortex-m7
CXX Compile Flags: -D __FPU_PRESENT=1 -DGEMMLOWP_ALLOW_SLOW_SCALAR_FALLBACK -DNDEBUG -DTF_LITE_DISABLE_X86_NEON -DTF_LITE_MCU_DEBUG_LOG -DTF_LITE_STATIC_MEMORY -MMD -O3 -Wall -Wextra -Wvla -Wno-missing-field-initializers -Wno-parentheses -Wno-sign-compare -Wno-strict-aliasing -Wno-type-limits -Wno-unused-but-set-variable -Wno-unused-parameter -Wno-unused-variable -Wno-write-strings -fdata-sections -ffunction-sections -fmessage-length=0 -fomit-frame-pointer -funsigned-char -fno-builtin -fno-delete-null-pointer-checks -fno-exceptions -fno-unwind-tables -mfloat-abi=hard -mlittle-endian -mthumb -mno-unaligned-access -nostdlib -std=c++11 -std=gnu++11 -fno-rtti -fpermissive -DARM_CMSIS_NN_M7 -DARM_MATH_CM7 -mcpu=cortex-m7 -mfpu=fpv5-sp-d16 -mtune=cortex-m7

BIN
src/libtf/cortex-m7/libtf.a Normal file

Binary file not shown.

View File

@ -0,0 +1,61 @@
/* This file is part of the OpenMV project.
* Copyright (c) 2013-2019 Ibrahim Abdelkader <iabdalkader@openmv.io> & Kwabena W. Agyeman <kwagyeman@openmv.io>
* This work is licensed under the MIT license, see the file LICENSE for details.
*/
#ifndef __LIBTF_H
#define __LIBTF_H
#ifdef __cplusplus
extern "C" {
#endif
// Call this first to get the shape of the model input.
// Returns 0 on success and 1 on failure.
// Errors are printed to stdout.
int libtf_get_input_data_hwc(const unsigned char *model_data, // TensorFlow Lite binary model (8-bit quant).
unsigned char *tensor_arena, // As big as you can make it scratch buffer.
const unsigned int tensor_arena_size, // Size of the above scratch buffer.
unsigned int *input_height, // Height for the model.
unsigned int *input_width, // Width for the model.
unsigned int *input_channels); // Channels for the model (1 for grayscale8 and 3 for rgb888).
// Call this second to get the shape of the model output.
// Returns 0 on success and 1 on failure.
// Errors are printed to stdout.
int libtf_get_output_data_hwc(const unsigned char *model_data, // TensorFlow Lite binary model (8-bit quant).
unsigned char *tensor_arena, // As big as you can make it scratch buffer.
const unsigned int tensor_arena_size, // Size of the above scratch buffer.
unsigned int *output_height, // Height for the model.
unsigned int *output_width, // Width for the model.
unsigned int *output_channels); // Channels for the model (1 for grayscale8 and 3 for rgb888).
// Callback to populate the model input data byte array (laid out in [height][width][channel] order).
typedef void (*libtf_input_data_callback_t)(void *callback_data,
unsigned char *model_input,
const unsigned int input_height,
const unsigned int input_width,
const unsigned int input_channels);
// Callback to use the model output data byte array (laid out in [height][width][channel] order).
typedef void (*libtf_output_data_callback_t)(void *callback_data,
unsigned char *model_output,
const unsigned int output_height,
const unsigned int output_width,
const unsigned int output_channels);
// Returns 0 on success and 1 on failure.
// Errors are printed to stdout.
int libtf_invoke(const unsigned char *model_data, // TensorFlow Lite binary model (8-bit quant).
unsigned char *tensor_arena, // As big as you can make it scratch buffer.
const unsigned int tensor_arena_size, // Size of the above scratch buffer.
libtf_input_data_callback_t input_callback, // Callback to populate the model input data byte array.
void *input_callback_data, // User data structure passed to input callback.
libtf_output_data_callback_t output_callback, // Callback to use the model output data byte array.
void *output_callback_data); // User data structure passed to output callback.
#ifdef __cplusplus
}
#endif
#endif // __LIBTF_H

Binary file not shown.

View File

@ -0,0 +1,27 @@
/* Copyright 2019 The TensorFlow Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
// This is a standard TensorFlow Lite model file that has been converted into a
// C data array, so it can be easily compiled into a binary for devices that
// don't have a file system. It was created using the command:
// xxd -i person_detect.tflite > person_detect_model_data.cc
#ifndef TENSORFLOW_LITE_EXPERIMENTAL_MICRO_EXAMPLES_PERSON_DETECTION_PERSON_DETECT_MODEL_DATA_H_
#define TENSORFLOW_LITE_EXPERIMENTAL_MICRO_EXAMPLES_PERSON_DETECTION_PERSON_DETECT_MODEL_DATA_H_
extern const unsigned char g_person_detect_model_data[];
extern const int g_person_detect_model_data_len;
#endif // TENSORFLOW_LITE_EXPERIMENTAL_MICRO_EXAMPLES_PERSON_DETECTION_PERSON_DETECT_MODEL_DATA_H_

@ -1 +1 @@
Subproject commit 00069f12cf48d12f6a84c242cc33966008637dfa
Subproject commit 452495f6c51233991eed7ac6b0f9ecd8011892ca

View File

@ -105,6 +105,7 @@ SRCS += $(addprefix py/, \
py_winc.c \
py_cpufreq.c \
py_nn.c \
py_tf.c \
)
OBJS = $(addprefix $(BUILD)/, $(SRCS:.c=.o))

View File

@ -83,4 +83,7 @@
// Enable LENET (200+ KB).
//#define IMLIB_ENABLE_LENET
// Enable Tensor Flow
//#define IMLIB_ENABLE_TF
#endif //__IMLIB_CONFIG_H__

View File

@ -128,6 +128,9 @@
// Enable LENET (200+ KB).
//#define IMLIB_ENABLE_LENET
// Enable Tensor Flow
//#define IMLIB_ENABLE_TF
// Enable FAST (20+ KBs).
//#define IMLIB_ENABLE_FAST

View File

@ -134,6 +134,9 @@
// Enable CMSIS NN
#define IMLIB_ENABLE_CNN
// Enable Tensor Flow
#define IMLIB_ENABLE_TF
// Enable FAST (20+ KBs).
#define IMLIB_ENABLE_FAST

View File

@ -69,7 +69,7 @@
#define OMV_MSC_BUF_SIZE (2K) // USB MSC bot data
#define OMV_VFS_BUF_SIZE (1K) // VFS sturct + FATFS file buffer (624 bytes)
#define OMV_FFS_BUF_SIZE (32K) // Flash filesystem cache
#define OMV_JPEG_BUF_SIZE (23 * 1024) // IDE JPEG buffer (header + data).
#define OMV_JPEG_BUF_SIZE (22 * 1024) // IDE JPEG buffer (header + data).
#define OMV_BOOT_ORIGIN 0x08000000
#define OMV_BOOT_LENGTH 32K

View File

@ -134,6 +134,9 @@
// Enable CMSIS NN
#define IMLIB_ENABLE_CNN
// Enable Tensor Flow
#define IMLIB_ENABLE_TF
// Enable FAST (20+ KBs).
#define IMLIB_ENABLE_FAST

View File

@ -72,8 +72,8 @@
#define OMV_FB_SIZE (400K) // FB memory: header + VGA/GS image
#define OMV_FB_ALLOC_SIZE (96K) // minimum fb alloc size
#define OMV_STACK_SIZE (7K)
#define OMV_HEAP_SIZE (240K)
#define OMV_STACK_SIZE (8K)
#define OMV_HEAP_SIZE (237K)
#define OMV_LINE_BUF_SIZE (3K) // Image line buffer round(640 * 2BPP * 2 buffers).
#define OMV_MSC_BUF_SIZE (12K) // USB MSC bot data

View File

@ -134,6 +134,9 @@
// Enable CMSIS NN
#define IMLIB_ENABLE_CNN
// Enable Tensor Flow
#define IMLIB_ENABLE_TF
// Enable FAST (20+ KBs).
#define IMLIB_ENABLE_FAST

View File

@ -75,8 +75,8 @@
#define OMV_FB_SIZE (30M) // FB memory: header + VGA/GS image
#define OMV_FB_ALLOC_SIZE (1M) // minimum fb alloc size
#define OMV_STACK_SIZE (7K)
#define OMV_HEAP_SIZE (240K)
#define OMV_STACK_SIZE (8K)
#define OMV_HEAP_SIZE (237K)
#define OMV_LINE_BUF_SIZE (11K) // Image line buffer round(2592 * 2BPP * 2 buffers).
#define OMV_MSC_BUF_SIZE (12K) // USB MSC bot data

443
src/omv/py/py_tf.c Normal file
View File

@ -0,0 +1,443 @@
/* This file is part of the OpenMV project.
* Copyright (c) 2013-2019 Ibrahim Abdelkader <iabdalkader@openmv.io> & Kwabena W. Agyeman <kwagyeman@openmv.io>
* This work is licensed under the MIT license, see the file LICENSE for details.
*/
#include <mp.h>
#include "py_helper.h"
#include "ff_wrapper.h"
#include "libtf.h"
#include "libtf_person_detect_model_data.h"
#ifdef IMLIB_ENABLE_TF
// TF Model Object
typedef struct py_tf_model_obj {
mp_obj_base_t base;
unsigned char *model_data;
unsigned int model_data_len, height, width, channels;
} py_tf_model_obj_t;
STATIC void py_tf_model_print(const mp_print_t *print, mp_obj_t self_in, mp_print_kind_t kind)
{
py_tf_model_obj_t *self = self_in;
mp_printf(print,
"{\"len\":%d, \"height\":%d, \"width\":%d, \"channels\":%d}",
self->model_data_len,
self->height,
self->width,
self->channels);
}
// TF Classification Object
#define py_tf_classification_obj_size 5
typedef struct py_tf_classification_obj {
mp_obj_base_t base;
mp_obj_t x, y, w, h, output;
} py_tf_classification_obj_t;
STATIC void py_tf_classification_print(const mp_print_t *print, mp_obj_t self_in, mp_print_kind_t kind)
{
py_tf_classification_obj_t *self = self_in;
mp_printf(print,
"{\"x\":%d, \"y\":%d, \"w\":%d, \"h\":%d, \"output\":",
mp_obj_get_int(self->x),
mp_obj_get_int(self->y),
mp_obj_get_int(self->w),
mp_obj_get_int(self->h));
mp_obj_print_helper(print, self->output, kind);
mp_printf(print, "}");
}
STATIC mp_obj_t py_tf_classification_subscr(mp_obj_t self_in, mp_obj_t index, mp_obj_t value)
{
if (value == MP_OBJ_SENTINEL) { // load
py_tf_classification_obj_t *self = self_in;
if (MP_OBJ_IS_TYPE(index, &mp_type_slice)) {
mp_bound_slice_t slice;
if (!mp_seq_get_fast_slice_indexes(py_tf_classification_obj_size, index, &slice)) {
nlr_raise(mp_obj_new_exception_msg(&mp_type_OSError, "only slices with step=1 (aka None) are supported"));
}
mp_obj_tuple_t *result = mp_obj_new_tuple(slice.stop - slice.start, NULL);
mp_seq_copy(result->items, &(self->x) + slice.start, result->len, mp_obj_t);
return result;
}
switch (mp_get_index(self->base.type, py_tf_classification_obj_size, index, false)) {
case 0: return self->x;
case 1: return self->y;
case 2: return self->w;
case 3: return self->h;
case 4: return self->output;
}
}
return MP_OBJ_NULL; // op not supported
}
mp_obj_t py_tf_classification_rect(mp_obj_t self_in)
{
return mp_obj_new_tuple(4, (mp_obj_t []) {((py_tf_classification_obj_t *) self_in)->x,
((py_tf_classification_obj_t *) self_in)->y,
((py_tf_classification_obj_t *) self_in)->w,
((py_tf_classification_obj_t *) self_in)->h});
}
mp_obj_t py_tf_classification_x(mp_obj_t self_in) { return ((py_tf_classification_obj_t *) self_in)->x; }
mp_obj_t py_tf_classification_y(mp_obj_t self_in) { return ((py_tf_classification_obj_t *) self_in)->y; }
mp_obj_t py_tf_classification_w(mp_obj_t self_in) { return ((py_tf_classification_obj_t *) self_in)->w; }
mp_obj_t py_tf_classification_h(mp_obj_t self_in) { return ((py_tf_classification_obj_t *) self_in)->h; }
mp_obj_t py_tf_classification_output(mp_obj_t self_in) { return ((py_tf_classification_obj_t *) self_in)->output; }
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_classification_rect_obj, py_tf_classification_rect);
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_classification_x_obj, py_tf_classification_x);
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_classification_y_obj, py_tf_classification_y);
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_classification_w_obj, py_tf_classification_w);
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_classification_h_obj, py_tf_classification_h);
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_classification_output_obj, py_tf_classification_output);
STATIC const mp_rom_map_elem_t py_tf_classification_locals_dict_table[] = {
{ MP_ROM_QSTR(MP_QSTR_rect), MP_ROM_PTR(&py_tf_classification_rect_obj) },
{ MP_ROM_QSTR(MP_QSTR_x), MP_ROM_PTR(&py_tf_classification_x_obj) },
{ MP_ROM_QSTR(MP_QSTR_y), MP_ROM_PTR(&py_tf_classification_y_obj) },
{ MP_ROM_QSTR(MP_QSTR_w), MP_ROM_PTR(&py_tf_classification_w_obj) },
{ MP_ROM_QSTR(MP_QSTR_h), MP_ROM_PTR(&py_tf_classification_h_obj) },
{ MP_ROM_QSTR(MP_QSTR_output), MP_ROM_PTR(&py_tf_classification_output_obj) }
};
STATIC MP_DEFINE_CONST_DICT(py_tf_classification_locals_dict, py_tf_classification_locals_dict_table);
static const mp_obj_type_t py_tf_classification_type = {
{ &mp_type_type },
.name = MP_QSTR_tf_classification,
.print = py_tf_classification_print,
.subscr = py_tf_classification_subscr,
.locals_dict = (mp_obj_t) &py_tf_classification_locals_dict
};
static const mp_obj_type_t py_tf_model_type;
typedef struct py_tf_input_data_callback_data {
image_t *img;
rectangle_t *roi;
} py_tf_input_data_callback_data_t;
STATIC void py_tf_input_data_callback(void *callback_data,
unsigned char *model_input,
const unsigned int input_height,
const unsigned int input_width,
const unsigned int input_channels)
{
py_tf_input_data_callback_data_t *arg = (py_tf_input_data_callback_data_t *) callback_data;
float xscale = input_width / ((float) arg->roi->w);
float yscale = input_height / ((float) arg->roi->h);
// MAX == KeepAspectRationByExpanding - MIN == KeepAspectRatio
float scale = IM_MAX(xscale, yscale), scale_inv = 1 / scale;
float x_offset = ((arg->roi->w * scale) - input_width) / 2;
float y_offset = ((arg->roi->h * scale) - input_height) / 2;
switch (arg->img->bpp) {
case IMAGE_BPP_BINARY: {
for (int y = 0, yy = input_height; y < yy; y++) {
uint32_t *row_ptr = IMAGE_COMPUTE_BINARY_PIXEL_ROW_PTR(arg->img, fast_floorf((y + y_offset) * scale_inv) + arg->roi->y);
int row = input_width * y;
for (int x = 0, xx = input_width; x < xx; x++) {
int pixel = IMAGE_GET_BINARY_PIXEL_FAST(row_ptr, fast_floorf((x + x_offset) * scale_inv) + arg->roi->x);
int index = row + x;
switch (input_channels) {
case 1: {
model_input[index] = COLOR_BINARY_TO_GRAYSCALE(pixel);
break;
}
case 3: {
int index_3 = index * 3;
pixel = COLOR_BINARY_TO_RGB565(pixel);
model_input[index_3 + 0] = COLOR_RGB565_TO_R8(pixel);
model_input[index_3 + 1] = COLOR_RGB565_TO_G8(pixel);
model_input[index_3 + 2] = COLOR_RGB565_TO_B8(pixel);
break;
}
default: {
break;
}
}
}
}
break;
}
case IMAGE_BPP_GRAYSCALE: {
for (int y = 0, yy = input_height; y < yy; y++) {
uint8_t *row_ptr = IMAGE_COMPUTE_GRAYSCALE_PIXEL_ROW_PTR(arg->img, fast_floorf((y + y_offset) * scale_inv) + arg->roi->y);
int row = input_width * y;
for (int x = 0, xx = input_width; x < xx; x++) {
int pixel = IMAGE_GET_GRAYSCALE_PIXEL_FAST(row_ptr, fast_floorf((x + x_offset) * scale_inv) + arg->roi->x);
int index = row + x;
switch (input_channels) {
case 1: {
model_input[index] = pixel;
break;
}
case 3: {
int index_3 = index * 3;
pixel = COLOR_GRAYSCALE_TO_RGB565(pixel);
model_input[index_3 + 0] = COLOR_RGB565_TO_R8(pixel);
model_input[index_3 + 1] = COLOR_RGB565_TO_G8(pixel);
model_input[index_3 + 2] = COLOR_RGB565_TO_B8(pixel);
break;
}
default: {
break;
}
}
}
}
break;
}
case IMAGE_BPP_RGB565: {
for (int y = 0, yy = input_height; y < yy; y++) {
uint16_t *row_ptr = IMAGE_COMPUTE_RGB565_PIXEL_ROW_PTR(arg->img, fast_floorf((y + y_offset) * scale_inv) + arg->roi->y);
int row = input_width * y;
for (int x = 0, xx = input_width; x < xx; x++) {
int pixel = IMAGE_GET_RGB565_PIXEL_FAST(row_ptr, fast_floorf((x + x_offset) * scale_inv) + arg->roi->x);
int index = row + x;
switch (input_channels) {
case 1: {
model_input[index] = COLOR_RGB565_TO_GRAYSCALE(pixel);
break;
}
case 3: {
int index_3 = index * 3;
model_input[index_3 + 0] = COLOR_RGB565_TO_R8(pixel);
model_input[index_3 + 1] = COLOR_RGB565_TO_G8(pixel);
model_input[index_3 + 2] = COLOR_RGB565_TO_B8(pixel);
break;
}
default: {
break;
}
}
}
}
break;
}
default: {
break;
}
}
}
typedef struct py_tf_classify_output_data_callback_data {
mp_obj_t out;
} py_tf_classify_output_data_callback_data_t;
STATIC void py_tf_classify_output_data_callback(void *callback_data,
unsigned char *model_output,
const unsigned int output_height,
const unsigned int output_width,
const unsigned int output_channels)
{
py_tf_classify_output_data_callback_data_t *arg = (py_tf_classify_output_data_callback_data_t *) callback_data;
PY_ASSERT_TRUE_MSG(output_height == 1, "Expected model output height to be 1!");
PY_ASSERT_TRUE_MSG(output_width == 1, "Expected model output width to be 1!");
arg->out = mp_obj_new_list(output_channels, NULL);
for (unsigned int i = 0; i < output_channels; i++) {
((mp_obj_list_t *) arg->out)->items[i] = mp_obj_new_float(model_output[i] / 255.0f);
}
}
STATIC mp_obj_t py_tf_classify(uint n_args, const mp_obj_t *args, mp_map_t *kw_args)
{
py_tf_model_obj_t *arg_model;
image_t *arg_img = py_helper_arg_to_image_mutable(args[1]);
rectangle_t roi;
py_helper_keyword_rectangle_roi(arg_img, n_args, args, 2, kw_args, &roi);
float arg_min_scale = py_helper_keyword_float(n_args, args, 4, kw_args, MP_OBJ_NEW_QSTR(MP_QSTR_min_scale), 1.0f);
PY_ASSERT_TRUE_MSG((0.0f < arg_min_scale) && (arg_min_scale <= 1.0f), "0 < min_scale <= 1");
float arg_scale_mul = py_helper_keyword_float(n_args, args, 5, kw_args, MP_OBJ_NEW_QSTR(MP_QSTR_scale_mul), 0.5f);
PY_ASSERT_TRUE_MSG((0.0f <= arg_scale_mul) && (arg_scale_mul < 1.0f), "0 <= scale_mul < 1");
float arg_x_overlap = py_helper_keyword_float(n_args, args, 6, kw_args, MP_OBJ_NEW_QSTR(MP_QSTR_x_overlap), 0.0f);
PY_ASSERT_TRUE_MSG(((0.0f <= arg_x_overlap) && (arg_x_overlap < 1.0f)) || (arg_x_overlap == -1.0f), "0 <= x_overlap < 1");
float arg_y_overlap = py_helper_keyword_float(n_args, args, 7, kw_args, MP_OBJ_NEW_QSTR(MP_QSTR_y_overlap), 0.0f);
PY_ASSERT_TRUE_MSG(((0.0f <= arg_y_overlap) && (arg_y_overlap < 1.0f)) || (arg_y_overlap == -1.0f), "0 <= y_overlap < 1");
fb_alloc_mark();
if (MP_OBJ_IS_TYPE(args[0], &py_tf_model_type)) {
arg_model = (py_tf_model_obj_t *) args[0];
} else {
const char *path = mp_obj_str_get_str(args[0]);
arg_model = m_new_obj(py_tf_model_obj_t);
arg_model->base.type = &py_tf_model_type;
if (!strcmp(path, "person_detection")) {
arg_model->model_data = (unsigned char *) g_person_detect_model_data;
arg_model->model_data_len = g_person_detect_model_data_len;
} else {
FIL fp;
file_read_open(&fp, path);
arg_model->model_data_len = f_size(&fp);
arg_model->model_data = fb_alloc(arg_model->model_data_len, FB_ALLOC_NO_HINT);
read_data(&fp, arg_model->model_data, arg_model->model_data_len);
file_close(&fp);
}
uint32_t tensor_arena_size;
uint8_t *tensor_arena = fb_alloc_all(&tensor_arena_size, FB_ALLOC_PREFER_SIZE);
PY_ASSERT_FALSE_MSG(libtf_get_input_data_hwc(arg_model->model_data,
tensor_arena,
tensor_arena_size,
&arg_model->height,
&arg_model->width,
&arg_model->channels),
"Unable to read model height, width, and channels!");
fb_free(); // Free just the fb_alloc_all() - not fb_alloc() above.
}
uint32_t tensor_arena_size;
uint8_t *tensor_arena = fb_alloc_all(&tensor_arena_size, FB_ALLOC_PREFER_SIZE);
mp_obj_t objects_list = mp_obj_new_list(0, NULL);
for (float scale = 1.0f; scale >= arg_min_scale; scale *= arg_scale_mul) {
// Either provide a subtle offset to center multiple detection windows or center the only detection window.
for (int y = roi.y + ((arg_y_overlap != -1.0f) ? (fmodf(roi.h, (roi.h * scale)) / 2.0f) : ((roi.h - (roi.h * scale)) / 2.0f));
// Finish when the detection window is outside of the ROI.
(y + (roi.h * scale)) <= (roi.y + roi.h);
// Step by an overlap amount accounting for scale or just terminate after one iteration.
y += ((arg_y_overlap != -1.0f) ? (roi.h * scale * (1.0f - arg_y_overlap)) : roi.h)) {
// Either provide a subtle offset to center multiple detection windows or center the only detection window.
for (int x = roi.x + ((arg_x_overlap != -1.0f) ? (fmodf(roi.w, (roi.w * scale)) / 2.0f) : ((roi.w - (roi.w * scale)) / 2.0f));
// Finish when the detection window is outside of the ROI.
(x + (roi.w * scale)) <= (roi.x + roi.w);
// Step by an overlap amount accounting for scale or just terminate after one iteration.
x += ((arg_x_overlap != -1.0f) ? (roi.w * scale * (1.0f - arg_x_overlap)) : roi.w)) {
rectangle_t new_roi;
rectangle_init(&new_roi, x, y, roi.w * scale, roi.h * scale);
if (rectangle_overlap(&roi, &new_roi)) { // Check if new_roi is null...
py_tf_input_data_callback_data_t py_tf_input_data_callback_data;
py_tf_input_data_callback_data.img = arg_img;
py_tf_input_data_callback_data.roi = &new_roi;
py_tf_classify_output_data_callback_data_t py_tf_classify_output_data_callback_data;
PY_ASSERT_FALSE_MSG(libtf_invoke(arg_model->model_data,
tensor_arena,
tensor_arena_size,
py_tf_input_data_callback,
&py_tf_input_data_callback_data,
py_tf_classify_output_data_callback,
&py_tf_classify_output_data_callback_data),
"Model classification failed!");
py_tf_classification_obj_t *o = m_new_obj(py_tf_classification_obj_t);
o->base.type = &py_tf_classification_type;
o->x = mp_obj_new_int(new_roi.x);
o->y = mp_obj_new_int(new_roi.y);
o->w = mp_obj_new_int(new_roi.w);
o->h = mp_obj_new_int(new_roi.h);
o->output = py_tf_classify_output_data_callback_data.out;
mp_obj_list_append(objects_list, o);
}
}
}
}
fb_alloc_free_till_mark();
return objects_list;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_KW(py_tf_classify_obj, 2, py_tf_classify);
mp_obj_t py_tf_len(mp_obj_t self_in) { return mp_obj_new_int(((py_tf_model_obj_t *) self_in)->model_data_len); }
mp_obj_t py_tf_height(mp_obj_t self_in) { return mp_obj_new_int(((py_tf_model_obj_t *) self_in)->height); }
mp_obj_t py_tf_width(mp_obj_t self_in) { return mp_obj_new_int(((py_tf_model_obj_t *) self_in)->width); }
mp_obj_t py_tf_channels(mp_obj_t self_in) { return mp_obj_new_int(((py_tf_model_obj_t *) self_in)->channels); }
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_len_obj, py_tf_len);
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_height_obj, py_tf_height);
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_width_obj, py_tf_width);
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_channels_obj, py_tf_channels);
STATIC const mp_rom_map_elem_t locals_dict_table[] = {
{ MP_ROM_QSTR(MP_QSTR_len), MP_ROM_PTR(&py_tf_len_obj) },
{ MP_ROM_QSTR(MP_QSTR_height), MP_ROM_PTR(&py_tf_height_obj) },
{ MP_ROM_QSTR(MP_QSTR_width), MP_ROM_PTR(&py_tf_width_obj) },
{ MP_ROM_QSTR(MP_QSTR_channels), MP_ROM_PTR(&py_tf_channels_obj) },
{ MP_ROM_QSTR(MP_QSTR_classify), MP_ROM_PTR(&py_tf_classify_obj) }
};
STATIC MP_DEFINE_CONST_DICT(locals_dict, locals_dict_table);
STATIC const mp_obj_type_t py_tf_model_type = {
{ &mp_type_type },
.name = MP_QSTR_tf_model,
.print = py_tf_model_print,
.locals_dict = (mp_obj_t) &locals_dict
};
STATIC mp_obj_t py_tf_load(mp_obj_t path_obj)
{
const char *path = mp_obj_str_get_str(path_obj);
py_tf_model_obj_t *tf_model = m_new_obj(py_tf_model_obj_t);
tf_model->base.type = &py_tf_model_type;
if (!strcmp(path, "person_detection")) {
tf_model->model_data = (unsigned char *) g_person_detect_model_data;
tf_model->model_data_len = g_person_detect_model_data_len;
} else {
FIL fp;
file_read_open(&fp, path);
tf_model->model_data_len = f_size(&fp);
tf_model->model_data = xalloc(tf_model->model_data_len);
read_data(&fp, tf_model->model_data, tf_model->model_data_len);
file_close(&fp);
}
fb_alloc_mark();
uint32_t tensor_arena_size;
uint8_t *tensor_arena = fb_alloc_all(&tensor_arena_size, FB_ALLOC_PREFER_SIZE);
PY_ASSERT_FALSE_MSG(libtf_get_input_data_hwc(tf_model->model_data,
tensor_arena,
tensor_arena_size,
&tf_model->height,
&tf_model->width,
&tf_model->channels),
"Unable to read model height, width, and channels!");
fb_alloc_free_till_mark();
return tf_model;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_1(py_tf_load_obj, py_tf_load);
#endif // IMLIB_ENABLE_TF
STATIC const mp_rom_map_elem_t globals_dict_table[] = {
{ MP_ROM_QSTR(MP_QSTR___name__), MP_OBJ_NEW_QSTR(MP_QSTR_tf) },
#ifdef IMLIB_ENABLE_TF
{ MP_ROM_QSTR(MP_QSTR_load), MP_ROM_PTR(&py_tf_load_obj) },
{ MP_ROM_QSTR(MP_QSTR_classify), MP_ROM_PTR(&py_tf_classify_obj) },
#else
{ MP_ROM_QSTR(MP_QSTR_load), MP_ROM_PTR(&py_func_unavailable_obj) },
{ MP_ROM_QSTR(MP_QSTR_classify), MP_ROM_PTR(&py_func_unavailable_obj) }
#endif // IMLIB_ENABLE_TF
};
STATIC MP_DEFINE_CONST_DICT(globals_dict, globals_dict_table);
const mp_obj_module_t tf_module = {
.base = { &mp_type_module },
.globals = (mp_obj_t) &globals_dict
};

View File

@ -1174,3 +1174,30 @@ Q(draw_ir)
// duplicate Q(scale)
Q(pixformat)
// duplciate Q(copy_to_fb)
// TensorFlow Module
Q(tf)
// duplicate Q(load)
Q(classify)
// Model Object
Q(tf_model)
// duplicate Q(len)
// duplicate Q(height)
// duplicate Q(width)
Q(channels)
// Classify
// duplicate Q(classify)
// duplicate Q(roi)
// duplicate Q(min_scale)
// duplicate Q(scale_mul)
// duplicate Q(x_overlap)
// duplicate Q(y_overlap)
// Class Object
Q(tf_classification)
// duplicate Q(x)
// duplicate Q(y)
// duplicate Q(w)
// duplicate Q(h)
Q(output)