Dear authors,
First of all, thank you for sharing this interesting work — the design of the Dynamic Template and Language module is very insightful and creative.
While reading through your codebase, I noticed that the dynamic template features — specifically aux_dict['attn_l2s'] from the backbone — do not seem to be used during inference. Instead, the template update mechanism used during inference appears to be different, and I could not find a description of it in the paper.
I would like to ask if this discrepancy might be due to a code version issue. If so, would it be possible to update the repository with the correct implementation? Alternatively, if I have misunderstood the intended design, please feel free to correct me.
Thank you in advance for your clarification and your excellent work!
Dear authors,
First of all, thank you for sharing this interesting work — the design of the Dynamic Template and Language module is very insightful and creative.
While reading through your codebase, I noticed that the dynamic template features — specifically aux_dict['attn_l2s'] from the backbone — do not seem to be used during inference. Instead, the template update mechanism used during inference appears to be different, and I could not find a description of it in the paper.
I would like to ask if this discrepancy might be due to a code version issue. If so, would it be possible to update the repository with the correct implementation? Alternatively, if I have misunderstood the intended design, please feel free to correct me.
Thank you in advance for your clarification and your excellent work!