chp

you can try changing pairs.th in the config file. the reference + source views of pairs.th aren't necessarily close as it is in the generalization training

https://github.com/aimagelab/meshed-memory-transformer/issues/16#issuecomment-617380393

Fixed the issue by changing Line 76 of tools/relation_test_net.py From ```python _ = checkpointer.load(cfg.MODEL.WEIGHT) ``` to ```python _ = checkpointer.load(cfg.MODEL.PRETRAINED_DETECTOR_CKPT) ``` seems like the pretrained weight were not loading

My understanding is that mip-nerf360 dataset is set to be a dataset type LLFF based on the below line. https://github.com/sunset1995/DirectVoxGO/blob/main/configs/nerf_unbounded/nerf_unbounded_default.py#L6 And LLFF datasets only leave on view out for validation/test....

I'm not saying that DVGO is not a good baseline. I'm just making sure the reported result is based on only one view because the dataset contains a 360 degree...