Sukjun Hwang

I suppose that's due to opencv environment conflict. Could you please elaborate?

It seems like the code is wrongly accessing the YTVIS data loader instead of OVIS. Could you check if OVIS configurations are being used correctly?

https://github.com/sukjunhwang/VITA/blob/f274a4eef298660b034b8b2a39a05278ad85e62a/vita/data/datasets/builtin.py#L41-L42 Could you please check if valid.json is under the folder named ```annotations```, instead of ```annotation```?

It's due to the sparse number of videos of video datasets. COCO gets augmented into videos, and used for training together.

I am very sorry for the late reply. Using V100 GPU with a ResNet-50 backbone, it runs at 46.1 FPS.

If using flattened images (say CxHxW -> CxHW), then I believe the positional encodings can act as an indicator for new rows, especially for fixed resolution settings. It'd be also...

FPS is measured by total frames in a validation dataset, divided by the total seconds it took to process the dataset. Total time gets scripted when an inference is completed,...

Hi @hoyeYang , The majority of the problems that occur at _clip_output.py_ are driven from either 1. too many instances captured or, 2. too long video sequence. Can you please...