An issue on demo.cpp

[MyChusterChuster]

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]

Thanks for your kind interpretation, learned a lot!

[MyChusterChuster]

Ive been working on a thesis and i want to know if you can let me some part of code of demo.cpp (specifically in object detection) putting GLP or Open Source license on it. Only if you want.

Thanks in advance. Cheers.