ali2500 / tarvis Goto Github PK

Hi
Thanks for your excellent job!
Did you provide visualization tool for Video Panoptic Segmentation on KITTI-STEP?

Hello, I am running your model on a new VOS dataset (like DAVIS and YoutubeVOS), but I don't know how to convert VOS dataset to json format. Can you tell me which script I should run?

Thanks for super awesome work! I have a quick question. TarVis model can operate in online method? I just want to use this model with real time application! Can I?

What an outstanding job! I'm curious to know how your group implement it, when will you release the code?

Hi,

Thanks for your open-source code! But I have the error below:

When I run "--model_dir my_first_tarvis_pretraining --cfg pretrain_resnet50_swin-tiny.yaml --amp --cfg.MODEL.BACKBONE ResNet50" on a single GPU, I got the below error, do you have any suggestion?
Num failed tries: 42
Num failed tries: 24
Num failed tries: 12
Num failed tries: 36
Num failed tries: 45
Num failed tries: 18
Num failed tries: 6
Num failed tries: 27
Num failed tries: 48
Num failed tries: 9
Num failed tries: 3

self.previous_clip_frame_indices = frame_indices[num_overlap_frames:].tolist()
self.previous_clip_logits = mask_logits[:, num_overlap_frames:]

If I'm understanding correctly, this should be the following since we intend to store last overlap_frames predictions?

self.previous_clip_frame_indices = frame_indices[- num_overlap_frames:].tolist()
self.previous_clip_logits = mask_logits[:, - num_overlap_frames:]

Please correct me if I'm wrong.
Thanks for sharing your amazing work!

Hi, authors

Thanks for your excellent work,

I would like to know where to download the pan_maps for coco, cityscapes and ade20k dataset, (/data/dataset_annotations/training/coco_panoptic/pan_maps/)

Thanks