Xueyan Zou comments

Results 49 comments of


                                            Xueyan Zou

How to train your own dataset

To construct JSON file please ask GPT4. Hhh

GPU Out of Memory Issue

Same suggestion, evaluating multiple images on a single image will cause: 1. Inaccurate evaluation (Because of padding). 2. OOM for GPU. I usually use 1 GPU for evaluation.

How can I use it for video track?

You can refer to the function here: https://github.com/UX-Decoder/Segment-Everything-Everywhere-All-At-Once/blob/7b2e76dbb17d0b7831c6813a921fe2bc8de22926/demo/seem/tasks/interactive.py#L187

the code doesn't work for me

What gpu env you are using?

Unable to reproduce open segmentation results for Pascal VOC

Thanks so much, I will take a look at the evaluation code.

About running the x-decoder demo

1. I tested my local environment, triton is not needed for X-Decoder/SEEM, as I didn't install it. 2. Which demo do you want to run? The current run_demo.sh are referring...

The shape mismatch in evaluation

Interesting, I have never met with this bug before. Have you ever change any code, including batchsize/image size and etc.?

The dataset type does not have the len() attribute

please install customized detectron2: git+https://github.com/MaureenZOU/detectron2-xyz.git

The dataset type does not have the len() attribute

pip install git+https://github.com/MaureenZOU/detectron2-xyz.git