mtcnn_facenet_cpp_tensorRT
mtcnn_facenet_cpp_tensorRT copied to clipboard
Embedding values are getting zero's
I am having known faces images in imgs folder with resolution 640 width and 480 height.For All of these images face is getting identified with valid output box. Some of these images with valid input data are being populated with 0 and -0, 128D output in doinference 'void FaceNetClassifier::doInference(float* inputData, float* output)' function.
Very similar images with same face and scene setting have values in output and sometimes zero's in output. Does the gender and race matters ?