Closed SanoPan0411 closed 6 years ago
yolo3 output layers are conv layers, and the output of these conv layers must be translated into intuitive predictions. The translation is implemented by computation of tensors, in yolo_eval
. And in yolo.py
there sess.run
runs the computation graph and get the predicitons.
So there is no yolo layer. and no need to use model.predict
.
I read the source code carefully,but I found there is no thing mentioned about yolo layer. I am new to yolov3,anyone can give a explanation of how yolo layer working? Thank you all.