Daniel Ji

> sorry, i find the mistakes made during preprocessing dataset..... Be carefull is pretty critical ! now i success running the code with cityscapes dataset, anyone who has problems can...

@Sibyl233 @ramdrop have you guys solved this problem yet? Could you give me some advice for that?

Assume it is not hard to implement such simple function, whatever by yourself or assisting with Codellama. You can just use our model to generate some prediction masks and generate...

Hi, @xyIsHere Do you have any progress on semantic segmentation?

Hi, @Isaachhh Appreciated your suggestions here. Actually, I print the shape of all `output_ids['hidden_states']`, casue it is a tuple. During the casual-decoder inference, they will generate 1024 items, and the...

I see. A nice discussion with you. Thank you again.

@gaozhitong I also encounter the same issue with you. Could you share your final solution about it? Thank you!

Does it affect the final performance a lot?

该数据集是我们团队发表于《机器智能研究MIR》上面关于医学结肠镜视频中息肉病灶区域分割的工作。如下是项目介绍： 结直肠癌作为世界第三大癌症群，慢慢演变成为工业化国家恶性死亡的第二大原因，严重威胁着人类的生命健康。大部分的结直肠癌是由腺瘤性息肉演变而来，若能通过早期结肠镜检查发现并切除癌前病变，可使得发病率大幅下降约30%左右，有效阻止结直肠癌的发生及发展。因此，内镜医师在检查时通过光学诊断方式准确地发现并判断息肉所在区域，将避免不必要的手术切除和病理检查，提高了结直肠镜筛查的成本-效益比。 本工作引入了领域内首个大规模视频息肉分割数据集SUN-SEG，其包含了1,106个视频片段以及158,690个视频帧，视频流畅度高达每秒30帧，并提供了具有密集型逐帧标注的分割标签、病理标签、视觉属性标签、弱监督标签等，为各类丰富的医学图像分析、处理任务提供了可能性，也启发了很多研究方向。目前该工作已经在Springer平台Open Access（网址：https://link.springer.com/article/10.1007/s11633-022-1371-y），详细内容可以参考原文。对应的数据集、基线模型、测评基准均已在GitHub开源（网址：https://github.com/GewelsJI/VPS）。

@linhandev hi could you process this issue in priority?