The effect of temporal scale on this model and hyperparameters.

[takfate]

In default BMN, the temporal scale is 100. if I want to generate 200*200 proposal map by feature with 200-length, need I change some hyperparameters ?

For instance, need I change the positive threshold of cls_pem_loss or three thresholds of reg_pem_loss ? Is there anything else that needs to be modified

[JJBOY]

As a matter of necessity, no hyper-parameters need to be changed. You only need to change temporal_scale from 100 to 200. But tuning the parameters can make all the difference in performance. In my experience, you can try to reduce num_sample or num_sample_perbin to reduce computation without degrading performance. You can also adjust the other hyperparameters to try.

[takfate]

As a matter of necessity, no hyper-parameters need to be changed. You only need to change temporal_scale from 100 to 200. But tuning the parameters can make all the difference in performance. In my experience, you can try to reduce num_sample or num_sample_perbin to reduce computation without degrading performance. You can also adjust the other hyperparameters to try.

Thanks a lot!

[xianguo-dev]

Hi! I'd like to know how to generate 200*200 proposal map by feature with 200-length? Although I have trained the TSN model, I don't know how to generate feature. Would you please share your code? Thanks a lot!