Paul
Results
2
comments of
Paul
注意一下路径，CDLA转COCO的目录里面的json要和paddledetection的配置对应起来。annotations.json
I've modified line 170 in model_net.py ```return x/torch.norm(x,2,1).view(x.shape[0], -1).repeat(1, self.feature_size)``` reached Recall ~35 python 3.6, cuda 8