Thank you very much for sharing the FVC code implementation on GitHub. I have a question while reading the code, and the issue mainly occurs during the training phase. I would like to inquire about the two-stage training approach used to train the model. In the first stage, it does not include multi-frame feature fusion, while in the second stage, it does include multi-frame feature fusion. It seems that the net.py file only contains the model for the first stage. May I ask if it's possible to provide the code that includes the multi-frame feature fusion module? My abilities are limited, and I would greatly appreciate it if you could provide the code for the second stage.
Thank you very much for sharing the FVC code implementation on GitHub. I have a question while reading the code, and the issue mainly occurs during the training phase. I would like to inquire about the two-stage training approach used to train the model. In the first stage, it does not include multi-frame feature fusion, while in the second stage, it does include multi-frame feature fusion. It seems that the net.py file only contains the model for the first stage. May I ask if it's possible to provide the code that includes the multi-frame feature fusion module? My abilities are limited, and I would greatly appreciate it if you could provide the code for the second stage.