In the original paper, the author used OKS (Object Keypoint Similarity) to evaluate the network when using COCO Dataset. I saw in simple-HRNet there's validation during training but in case I want to custom COCO dataset (adding my own data), then retrain again. After that I need to evaluate the performance on test videos, not the ones in training, is there any function to do so?
In the original paper, the author used OKS (Object Keypoint Similarity) to evaluate the network when using COCO Dataset. I saw in simple-HRNet there's validation during training but in case I want to custom COCO dataset (adding my own data), then retrain again. After that I need to evaluate the performance on test videos, not the ones in training, is there any function to do so?