Based on RTSPFaceDetection.ino,
how do I create the following add into FDPostProcess ?
- img = CV2.crop(frame)
- img = CV2.rgb2gray(img)
- img = CV2.resize(img, (48, 48))
- emo = NNemotiondetection.predict(img)
my emotion detection model is trained by
kaggle.com/rkuo2000/fer2013
Pretrained README
file : fer2013_cnn.h5 (1.26MB)
total params = 103,143
input_shape = (48,48,1)
test accuracy = 54.89%
( minimize model size, also reduce accuracy! )