Hi, thanks for your impressive work and for making the codebase publicly available. The paper and the released modules are very well organized, and they have been really helpful for my research. I truly appreciate the effort your team put into this project.
I have a few questions when trying to reproduce the experiments related to OverLoCK:
First, I’m trying to locate the experimental code for the OverLoCK model, but the project contains many files and I’m not sure whether the full training/inference pipeline is included.If the experimental code is not fully released, could you briefly describe how the SFM module is added on top of OverLoCK in your experiments?
Second, I noticed that the adaptive downsampling module does not include a channel-expansion Conv2d layer. In your implementation, do we need to manually add a Conv2d to increase the channel dimension after feeding features downstream, or is the module designed to work without channel expansion?
Thanks again for your time and for sharing the code. I appreciate any clarification you can provide.
Hi, thanks for your impressive work and for making the codebase publicly available. The paper and the released modules are very well organized, and they have been really helpful for my research. I truly appreciate the effort your team put into this project.
I have a few questions when trying to reproduce the experiments related to OverLoCK:
First, I’m trying to locate the experimental code for the OverLoCK model, but the project contains many files and I’m not sure whether the full training/inference pipeline is included.If the experimental code is not fully released, could you briefly describe how the SFM module is added on top of OverLoCK in your experiments?
Second, I noticed that the adaptive downsampling module does not include a channel-expansion Conv2d layer. In your implementation, do we need to manually add a Conv2d to increase the channel dimension after feeding features downstream, or is the module designed to work without channel expansion?
Thanks again for your time and for sharing the code. I appreciate any clarification you can provide.