Open ghost opened 4 years ago
hi,I want to train with hight resolution input such as 1024*1024. How can I design the network with hight resoulution input and inference fast. Could you give me some advice. THANKS.
Set subdivisions=batch and random=0 in yolo layer, if out of memory you can try yolov3 tiny.
Your can also divide the image into crops and train on each crop...
hi,I want to train with hight resolution input such as 1024*1024. How can I design the network with hight resoulution input and inference fast. Could you give me some advice. THANKS.