Ruijin Liu

@leoxiaobin Thanks for your amazing work, I am also interested in the tracking done by MXnet.

Hi, thanks for your interest! You could e-mail [email protected] to ask the company for use. I have all videos but I have no right to distribute them.

Thanks for your interest and affirmation! I'm also looking forward to seeing how it performs on more complex datasets. Feel free to pull requests after the code is public.

Thanks for your patience, codes have been public @jinfagang .

Hi, Yes, the multi-GPU training is not working. I might fix it in a future version. Thanks a lot!

@cmiller2air @Starboy-at-earth Sorry for the months of waiting, I have done the multi-GPU codes and will release them after some testing 🤣 @Starboy-at-earth CPU: AMD Ryzen 1900x GPU: 2x1080Ti DIMM:...

Hi， I suppose https://github.com/liuruijin17/LSTR/blob/c8fe59a1d8e1c456c982c9e954ee660ae9116b49/test/tusimple.py#L118 is the sentence you need.

Hi, @qklee-lz thanks for your interest! A1. I have fixed the multi-GPU codes and will release them. Q2. chunk_sizes: [16] means: use one GPU, put 16 images on it. After...

Hi, thanks for your interest! Emm, your computer is so good. Our batch16 only gets 420 on my PC. CPU: AMD Ryzen 1900x GPU: 2x1080Ti DIMM: 4x16 DDR4 3200MHz SSD:...

Hi, thanks for your interest! Please refer to #8 After training on your own lane datasets, the performance would be better.