search

dipika-singhania / ICC-Semi-Supervised-TAS

Iterative Contrast-Classify For Semi-supervised Temporal Action Segmentation
MIT License
9 stars 0 forks source link

Fail to reproduce the results for 50salads #2

Closed tqosu closed 2 years ago

tqosu commented 2 years ago

Hi Dipika,

I used your implement code and weights for the 50salads dataset. But the results were much worse than that in the paper.

Do you know how to fix this?

Implementation results 5% 41.7 & 36.1 & 25.7 & 38.9 & 55.0 10% 61.1 & 55.6 & 42.0 & 52.3 & 65.8

results in the paper 5% 52.9 49.0 36.6 45.6 61.3 10% 67.3 64.9 49.2 56.9 68.6

dipika-singhania commented 2 years ago

Please download the unsupervised checkpoint again. Also, update the runcodes semi-supervised scripts and then try to run the whole thing again. There was a hyper-parameter which was hard coded to constant value. It would be replicated.

tqosu commented 2 years ago

Hi Dipika,

Thanks.👍 I will have a try.

Best, Tieqiao

On Sat, Jun 18, 2022 at 3:46 AM dipika singhania @.***> wrote:

[This email originated from outside of OSU. Use caution with links and attachments.]

Closed #2 https://github.com/dipika-singhania/ICC-Semi-Supervised-TAS/issues/2 as completed.

— Reply to this email directly, view it on GitHub https://github.com/dipika-singhania/ICC-Semi-Supervised-TAS/issues/2#event-6832687256, or unsubscribe https://github.com/notifications/unsubscribe-auth/AVY6QGE2JBKNGCBRZ6IVYHLVPWSKBANCNFSM5YSL5TMA . You are receiving this because you authored the thread.Message ID: <dipika-singhania/ICC-Semi-Supervised-TAS/issue/2/issue_event/6832687256@ github.com>