search

JerryX1110 / RPCMVOS

[AAAI22 Oral] Reliable Propagation-Correction Modulation for Video Object Segmentation
MIT License
78 stars 11 forks source link

About multi-scale testing #2

Closed longmalongma closed 2 years ago

longmalongma commented 2 years ago

Hello, thank you very much for your great work. Can you provide your multi-scale test code?

JerryX1110 commented 2 years ago

Thanks! Due to the policy in the company, I can't share the code with you right now. After a code review in my company, I will release it asap. Specifically, we follow the multi-scale evaluation method as CFBI and directly average the probability map of several scales. Then, we use an argmax function for the final ensemble.

By the way, you can find that the multi-scale testing only brings a minor gain for our model. I think simply conducting an averaging operation is not an optimal way for the ensemble.

longmalongma commented 2 years ago

Ok, thanks! Looking forward to your code!