Unexpected results for pre-trained model

[StOnEGiggity]

Hi,

Thanks for your work~

I evaluate three different pre-trained models (ANS(rgb), OccAnt(depth) and OccAnt(depth) with challenge vision) with the noise-free setting. However, the SPL and Success seem almost zero. I guess there may be some mistakes in my own evaluation config. Is that possible to provide the corresponding evaluation script?

By the way, the model that I trained from scratch performs badly, either. If you have any suggestion, please tell me. Thank you very much~

[srama2512]

Hi @StOnEGiggity,

I'm afraid I need more information to suggest what steps you need to take. Could you provide me with the config, model paths, and the commands used for evaluation?