search

GitHubOfHyl97 / SkeAttnCLR

The Official PyTorch implementation of "Part Aware Contrastive Learning for Self-Supervised Action Recognition" in IJCAI 2023
14 stars 3 forks source link

how can I do the knn evaluation experiment #1

Open lbycan opened 1 year ago

lbycan commented 1 year ago

Hi! Thanks for your great work! I have a doubt that how can I use your code to do the KNN evaluation experiment. Is it should have T-SNE at the same time?

GitHubOfHyl97 commented 1 year ago

Hello! In this code, the KNN evaluattion is set in the validation part and is performed every few epoches during training. Besides, during the training process, the code saves the KNN results and model parameters of each validation. So, you can select the one with the best KNN results for T-SNE after training process.

---- Replied Message ---- | From | @.> | | Date | 06/26/2023 12:25 | | To | @.> | | Cc | @.***> | | Subject | [GitHubOfHyl97/SkeAttnCLR] how can I do the knn evaluation experiment (Issue #1) |

Hi! Thanks for your great work! I have a doubt that how can I use your code to do the KNN evaluation experiment. Is it should have T-SNE at the same time?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>

lbycan commented 1 year ago

Thanks for your reply! Yes, I know that you use KNN every 10 epochs from your paper. But if the KNN evaluation is performed every few epochs during training, is there a slow training speed? Actually, I just want to evaluate the latest backbone parameters by the KNN classifier. I will try to do it, Thank you!