Jinkun Cao issues

Results 18 issues of


                                            Jinkun Cao

add paper links to 'Human Pose Estimation'

add links to six papers

dataset config error when training on CARLA dataset

I tried to train esp with carla town01 dataset downloaded from the link you provided. As all .json files in the provided dataset are named "ma_datum_*.json", so I modified in...

Train without --masks flag

Hi, I tried to train a model without --masks on the default Youtube VIS dataset. I used all parameters as default except for decreasing the num_frames to 18 and num_queries...

delete unnecessary code

if self.return_intermediate is enabled, the line 117 - 119 are not necessary, whose function has been covered by line 112 - 113 already.

[Feature] Support OC-SORT for MOT

Support `Observation-Centric SORT: Rethinking SORT for Robust Multi-Object Tracking` ([official code](https://github.com/noahcao/OC_SORT), [arxiv](https://arxiv.org/abs/2203.14360)). The implementation has been verified on the MOT17-val set. [WIP]: 1. Polish the readme file at `configs/mot/ocsort/README.md` 2....

Experiments on more datasets?

Thanks for the excellent! I'm curious if you have tried this model on other popular datasets to study disentanglement, e.g., dSprites, SmallNORB, Cars3D, Shapes3D as benchmarked in [disentanglement_lib](https://github.com/google-research/disentanglement_lib). I tried...

overlap of masks in MOTS evaluation

I noticed a code [here](https://github.com/JonathonLuiten/TrackEval/blob/bcd03a6cc5f4fa0074da89d95e859fe77e264c3e/trackeval/datasets/mots_challenge.py#L267) not allowing the masks of different objects to have overlap. Is there any reason or insight to set like this？ Considering the recent instance segmentation...

Jinkun Cao