Hi, thank you for releasing the VSCD code and dataset.
I would like to reproduce the evaluation results using the provided eval.py. The README requires a trained model checkpoint (for example, runs/vscd_default/ckpt_best.pt), but I could not find a pretrained VSCD checkpoint in the repository, GitHub Releases, or the author's Hugging Face account.
Are there plans to release the official pretrained checkpoint and, if applicable, the exact configuration used to obtain the reported results? Having these files would make it possible to run evaluation directly and verify the benchmark results without retraining the full model.
Thank you for your help.
Hi, thank you for releasing the VSCD code and dataset.
I would like to reproduce the evaluation results using the provided eval.py. The README requires a trained model checkpoint (for example, runs/vscd_default/ckpt_best.pt), but I could not find a pretrained VSCD checkpoint in the repository, GitHub Releases, or the author's Hugging Face account.
Are there plans to release the official pretrained checkpoint and, if applicable, the exact configuration used to obtain the reported results? Having these files would make it possible to run evaluation directly and verify the benchmark results without retraining the full model.
Thank you for your help.