Closed JohannesEsslinger closed 2 years ago
Hi i am wondering why you used image norm for ImageNet even if you train from scratch for COCO?
https://github.com/open-mmlab/mmdetection/blob/master/configs/ssd/ssdlite_mobilenetv2_scratch_600e_coco.py
It is just a convention to normalize the image by the norm on ImageNet since it is large and the values should be representative and general.
Hi i am wondering why you used image norm for ImageNet even if you train from scratch for COCO?
https://github.com/open-mmlab/mmdetection/blob/master/configs/ssd/ssdlite_mobilenetv2_scratch_600e_coco.py