urbaneman

Results 7 comments of urbaneman

How does the code in the psconv.py file achieve the different dilation rates written in the paper?

> > How does the code in the psconv.py file achieve the differentdilation rates written in the paper? > > I think the different dilation rates are done through the...

where to find "supplementary" materials？

I have the same question，in paper Eq.8 > ". In the following experiments, we empirically set C to the average absolute size of AI-TOD and achieve the best performance. Moreover,...

Finetuning error: "Creating layer conv14_2_mbox_conf_flat" --> "Check failed: std::equal(top_shape.begin() + 1, top_shape.begin() + 4, shape.begin() + 1) "

I had this issue too? Did you solution this? Maybe, the lmdbs have error? But I'm not sure. I used my old lmdbs in the same Net and parameters, the...

既然是手机屏幕，我觉得可以把触屏安排上

嗯，想法很棒，想着想着头就秃了:joy:

How to run inference on single video?

@tarunmcom I added this feature in my own fork project, you can refer to it [support video infer](https://github.com/urbaneman/MOTIP/commit/f25ca760f588d3b7c89b43d646537c02dc35c726)

Abount CC3M Pseudo Annotations use

> As far as I know, they use an encryption method called zlib.crc32 in the file file_name. I think I've managed to get pseudo annotation results that match the dataset,...

关于真值中车道线outlier点的问题

是的，通过在视频上的可视化，能够发现很多标签的问题，1. 两条平行的车道线连在了一起，2. 车道线类别跳变，3. 车道线漏标（闪烁）。整体标注质量不太行，2D和3D的车道线精度都不太行。 我现在的想法是基于跟踪的方式和车道线建模的方式做筛选，如果有人力的话就补标，没有的话直接过滤对应的片段。