How edge-impulse-sdk is inferencing object detection without Registering TFLite_Detection_PostProcess()

Ramson · April 27, 2022, 7:08am

Hi,
I have trained a object detection model by selecting MobileNetV2 SSD FPN-Lite 320x320. The model has TFLite_Detection_PostProcess layer at the end of the model, which outputs 4 outputs as shown in the image below:

I exported the project as c++ library without compiling with EON compiler. As i looked into the sdk, the allopsresolver didn’t have Detection_PostProcess operator, it was commented out though and EI_CLASSIFIER_ENABLE_DETECTION_POSTPROCESS_OP was not defined as well.

I have seen before that if a operator is missing in the resolver, error would through out that operator is missing, In this case i wonder how it works? Please can anyone explain how it is done.

Thanks & Regards,
Ramson Jehu K

dansitu · April 27, 2022, 7:45am

Hi Ramson,

Thanks for your question! If you’re not using EON you’ll need to add EI_CLASSIFIER_ENABLE_DETECTION_POSTPROCESS_OP to your compiler flags to ensure the op is available. There’s an example of this in our Example Standalone Inferencing repos:

Bare metal:

github.com

edgeimpulse/example-standalone-inferencing/blob/master/Makefile#L21

      
        
            
            
# Search path for header files (current directory)
            CFLAGS += -I.
            
            
# C and C++ Compiler flags
            CFLAGS += -Wall						# Include all warnings
            CFLAGS += -g						# Generate GDB debugger information
            CFLAGS += -Wno-strict-aliasing		# Disable warnings about strict aliasing
            CFLAGS += -Os						# Optimize for size
            CFLAGS += -DNDEBUG					# Disable assert() macro
            CFLAGS += -DEI_CLASSIFIER_ENABLE_DETECTION_POSTPROCESS_OP	# Add TFLite_Detection_PostProcess operation
            
            
# C++ only compiler flags
            CXXFLAGS += -std=c++14				# Use C++14 standard
            
            
# Linker flags
            LDFLAGS += -lm 						# Link to math.h
            LDFLAGS += -lstdc++					# Link to stdc++.h
            
            
# Include C source code for required libraries
            CSOURCES += $(wildcard edge-impulse-sdk/CMSIS/DSP/Source/TransformFunctions/*.c) \

Linux:

github.com

edgeimpulse/example-standalone-inferencing-linux/blob/master/Makefile#L13

      
        
            UNAME_S := $(shell uname -s)
            
            
CFLAGS +=  -Wall -g -Wno-strict-aliasing
            CFLAGS += -I.
            CFLAGS += -Isource
            CFLAGS += -Imodel-parameters
            CFLAGS += -Itflite-model
            CFLAGS += -Ithird_party/
            CFLAGS += -Os
            CFLAGS += -DNDEBUG
            CFLAGS += -DEI_CLASSIFIER_ENABLE_DETECTION_POSTPROCESS_OP
            CFLAGS += -g
            CXXFLAGS += -std=c++14
            LDFLAGS += -lm -lstdc++
            
            
CSOURCES = $(wildcard edge-impulse-sdk/CMSIS/DSP/Source/TransformFunctions/*.c) $(wildcard edge-impulse-sdk/CMSIS/DSP/Source/CommonTables/*.c) $(wildcard edge-impulse-sdk/CMSIS/DSP/Source/BasicMathFunctions/*.c) $(wildcard edge-impulse-sdk/CMSIS/DSP/Source/ComplexMathFunctions/*.c) $(wildcard edge-impulse-sdk/CMSIS/DSP/Source/FastMathFunctions/*.c) $(wildcard edge-impulse-sdk/CMSIS/DSP/Source/SupportFunctions/*.c) $(wildcard edge-impulse-sdk/CMSIS/DSP/Source/MatrixFunctions/*.c) $(wildcard edge-impulse-sdk/CMSIS/DSP/Source/StatisticsFunctions/*.c)
            CXXSOURCES = $(wildcard tflite-model/*.cpp) $(wildcard edge-impulse-sdk/dsp/kissfft/*.cpp) $(wildcard edge-impulse-sdk/dsp/dct/*.cpp) $(wildcard ./edge-impulse-sdk/dsp/memory.cpp) $(wildcard edge-impulse-sdk/porting/posix/*.c*) $(wildcard edge-impulse-sdk/porting/mingw32/*.c*)
            CCSOURCES =
            
            
ifeq (${USE_FULL_TFLITE},1)
            CFLAGS += -DEI_CLASSIFIER_USE_FULL_TFLITE=1

Thank you for using Edge Impulse!

Warmly,
Dan

Ramson · April 27, 2022, 8:40am

Thanks Dan. You are amazing.