Hi, I am writing a tensorflow 2 version of the Seamless Single Shot. I found it is not accurate as I expected and may be due to that I did not pretrain my network. I am wondering how did you go through the pretrain? There are several datasets in COCO, which one did you use?
Hi, I am writing a tensorflow 2 version of the Seamless Single Shot. I found it is not accurate as I expected and may be due to that I did not pretrain my network. I am wondering how did you go through the pretrain? There are several datasets in COCO, which one did you use?