finnickniu / tensorflow_object_detection_tflite Goto Github PK

CMakeFiles/demo.dir/demo.cpp.o: In function test()': demo.cpp:(.text+0x862): undefined reference to tflite::InterpreterBuilder::InterpreterBuilder(tflite::FlatBufferModel const&, tflite::OpResolver const&)'
demo.cpp:(.text+0x87b): undefined reference to tflite::InterpreterBuilder::operator()(std::unique_ptr<tflite::Interpreter, std::default_delete<tflite::Interpreter> >*)' demo.cpp:(.text+0x88a): undefined reference to tflite::InterpreterBuilder::~InterpreterBuilder()'
demo.cpp:(.text+0xaf9): undefined reference to tflite::Interpreter::AllocateTensors()' demo.cpp:(.text+0xbd9): undefined reference to tflite::Interpreter::SetAllowFp16PrecisionForFp32(bool)'
demo.cpp:(.text+0xbf5): undefined reference to tflite::Interpreter::SetNumThreads(int)' demo.cpp:(.text+0xc0c): undefined reference to tflite::Interpreter::Invoke()'
demo.cpp:(.text+0x1521): undefined reference to tflite::InterpreterBuilder::~InterpreterBuilder()' CMakeFiles/demo.dir/demo.cpp.o: In function std::default_deletetflite::Interpreter::operator()(tflite::Interpreter*) const':
demo.cpp:(.text.ZNKSt14default_deleteIN6tflite11InterpreterEEclEPS1[ZNKSt14default_deleteIN6tflite11InterpreterEEclEPS1]+0x1e): undefined reference to `tflite::Interpreter::~Interpreter()'
collect2: error: ld returned 1 exit status
CMakeFiles/demo.dir/build.make:111: recipe for target 'demo' failed
make[2]: *** [demo] Error 1
CMakeFiles/Makefile2:67: recipe for target 'CMakeFiles/demo.dir/all' failed
make[1]: *** [CMakeFiles/demo.dir/all] Error 2
Makefile:83: recipe for target 'all' failed
make: *** [all] Error 2

I'm not able to resolve this error. Can someone help me?

First of all, thanks for sharing the code publicly, thanks for share a c++ code with tensorflow. I was working on codes with OpenCV, Tensorflow and many vision libs and i saw in your code the score is focusing on num_detections instead of "score outputs".

In the line
float score = expit(nums[count]); // How has this to be done?

where nums came from num_detections, this should have been from output scores like:
scores = interpreter->tensor(interpreter->outputs()[2]);
auto scrs = scores->data.f;

In this position, interpreter gives you "the score outputs" and then this line would change to:
float score = expit(scrs[count]);

After all these changes, we can see how expit function has make more sense values. I hope this helps you.

@finnickniu I retrained an SSD Mobilenet model and converted it to tflite. I'm able to get accurate detections using the python code however while inferencing with C++ code, the score seems to be incorrect and therefore I'm getting a lot of false detections. Also the score seems to be always 0.9995.
In the python code the score is taken from the 3rd output tensor, but in the C++ code it is extracted by calling the expit function with nums[count] as argument. Why is this so ?

Also, I'm only trying to detect one class.

A quick reply will be really appreciated as I'm in a race against time. Thanks

[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Core was generated by `./demo'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0 __memcpy_avx_unaligned () at ../sysdeps/x86_64/multiarch/memcpy-avx-unaligned.S:238
238 ../sysdeps/x86_64/multiarch/memcpy-avx-unaligned.S: No such file or directory.
[Current thread is 1 (Thread 0x7fadc938c9c0 (LWP 27232))]

Hi,
I was working with TFLite. I tried the infernece code for the tflite model in your repo. It is working good.
But when i run the code with the SSD mobilenet V2 tflite model, i get wrong classes and also boxes make no sense...Is this something you noticed?

Can you please help me?

I convert the model using following commands.
python object_detection/export_tflite_ssd_graph.py
--pipeline_config_path=$CONFIG_FILE
--trained_checkpoint_prefix=$CHKPT_DIR
--output_directory=$MODEL_DIR
--add_postprocessing_op=true

tflite_convert --graph_def_file $MODEL_DIR/tflite_graph.pb --output_file $MODEL_DIR/detect.tflite
--input_arrays=normalized_input_image_tensor
--output_arrays='TFLite_Detection_PostProcess','TFLite_Detection_PostProcess:1','TFLite_Detection_PostProcess:2','TFLite_Detection_PostProcess:3'
--input_shapes=1,300,300,3 --inference_type=QUANTIZED_UINT8
--mean_values=128 --std_dev_values=128
--change_concat_input_ranges=false --allow_custom_ops
--default_ranges_min=0 --default_ranges_max=255