New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
train No.3 #154
Comments
Hi! First of all check if provided pre-trained model works for validation. This will say if validation works. Then check output file with prediction results from your checkpoint. If it is empty, then such error may occur. So, if it is empty, try to visualize the avg_heatmaps to see if anything was detected. Loss curves are here: #10, yours look reasonable. |
Thank you for your prompt reply. Thank you for your advice. I'll have a try. I also want to ask you two question. Firstly, the key points I want to set now is 20, and I want to add a waist node. I would like to ask you if I can directly use the COCO data set? However, the COCO data set only annotates 18 key points. If I want to set 20 key points, do I need to annotate the COCO data set again? |
Yes, you should label somehow waist for the persons, either manually or compute from existing ones. Keypoints can be detected after the third step (you see, there is validation step during training, so after some initial iterations, e.g. 5000, keypoints are detected, try to visualize the heatmaps). |
|
Hello! I want to ask two questions. Firstly, I tried to train the No.3 (--from-mobilenet)and train 285000 iters. However, I did not detect any points and lines when I ran the Demo with it. Do you know what the problem is?Is it because I didn't train for step four or five? Secondly, Can you tell me how much you lost at the end of your training? I want to know the approximate loss when we can stop training. I trained 285000 iters, my losses are as follows:
The text was updated successfully, but these errors were encountered: