dynamic crop function returns a tensor with shape (64, 112, 112, 5) while model expects input with shape (64, 224, 224, 5). Do you use any other processing technique before feed data into model? Currently I'm using cv2.resize to resize each frame from 112 to 224. Is it a right approach? In the paper, did you use dynamic crop to achieve 87% accuracy?
dynamic crop function returns a tensor with shape (64, 112, 112, 5) while model expects input with shape (64, 224, 224, 5). Do you use any other processing technique before feed data into model? Currently I'm using
cv2.resize
to resize each frame from 112 to 224. Is it a right approach? In the paper, did you use dynamic crop to achieve 87% accuracy?