yangliu14 / open-world-tracking Goto Github PK

Hello, By disturb you again. May I ask if the final experiment result of OWTB comes from the config "Detectron2-OWT/Configs/New_BASELINES/MASK_RCNN_R_101_FPN_400EP_LSJ_OWT.PY" ?

Hello, Thanks for the excellent work. Do you plan to release pre-trained models? If yes, when can we expect them to be released.

Hello, Thanks for the excellent work,looking forward to your response

Could you provide the results of the proposal generation and tracking? Thanks!

I just find the original TAO dataset in this project wesite https://openworldtracking.github.io/. How can i download TAO-OW dataset?
我只在官网https://openworldtracking.github.io/找到了TAO原本的数据集连接，没找到TAO-OW的连接，请问我该如何下载。

Hi, sorry to bother you again. I'm commited to solving the existing problems in OWT proposed by your paper. As mentioned in your paper, the baseline uses both optical flow and re-id similarity scoring and combines these into final tracks using tracklet merging. But I can't find the code to generate proposals' motion with optical flow. Looking forward to your reply. Thank you!

Hello, since i have checked coco_id2tao_id.json，i found that there are only 78 keys rather than 80 in the json file. And i try to use coco annotations to convert coco id to tao id, but i failed for some id are not in coco annotations. Could you tell me how to use the coco_id2tao_id.json please ! Thank you !

Hello, we want to follow your amazing work. And I found that the full code for "Association Similarity"has not been published yet. Could you publish this full code please, thank you.

I found some errors when running the generating proposals code. When running gen_proposals.py, the input path should be /data/TAO/frames, not including /val, or the code could not find any dataset. Besides, I got the Attribute Errors that cannot find all the ‘bg_score’, ‘objectness’, ‘embeddings’ fields in the given instances. Back to code, it is in the owt_utils.py, at about 40-50 lines. Could you please tell me how to fix it? Thank you.

Hello, we are happy to follow this good job. We met some problems with the training flow. Is the model pre-trained on coco dataset first and then finetuned on the tao_ow_train dataset?

Hello, after reading this excellent paper, I want to try to run the code, could you provide the weights file? Thank you very much!