GAO BIN
GAO BIN
I would like to know the details of result from your paper(Table 4.). For VID dateset, ur DFF got 73.1 mAP and 20.25fps, Could u pls give me more details...
> > > 您好，我用自己数据训练，后期boundery_bce_loss和boundery_dice_loss为0，您知道是什么原因吗？ > > > > > > 不太清楚，不过我建议可以可视化一下你的detail gt，是不是阈值设置的问题导致detail gt最后空了？毕竟阈值0.1是cityscapes上用的；仅仅是个建议哈，不一定就是真实原因，可能需要一点点排查~ > > 好的，谢谢哈 @sunvina 您好，请问下你找到原因了吗？同遇到这个问题
Hi, I have same problem with u, Could u pls update the dataset of voc2007 proposals? Thx
模型不收敛 loss越来越大的问题，需要在dict.txt里面创建一个id给nulll用，建议如下形式： ''' 13 0 y 1 m 2 d 3 1 4 2 5 3 6 4 7 5 8 6 9 7 10 8 11 9 12 0...
@daming98 github会显示空白，所以我加了两个@符号，请自己去掉就可以了。
你好，我在自己项目的数据集train的，大概800m
> The voxel task + rendering code is at https://github.com/d4l3k/torchdrive/blob/main/torchdrive/tasks/voxel.py > > The custom raymarcher that I'm using to render depth + semantic classes is at https://github.com/d4l3k/torchdrive/blob/main/torchdrive/raymarcher.py#L10 > > For...
@d4l3k what i want to do is that since my network produce occupancy grid, then i render depths from the occupancy grid, we can compare render depths with our depth...
@d4l3k hhh, got it, I didn’t read the code carefully earlier.