NExT VideoLLM Team
NExT VideoLLM Team
> Hi @sH1cHEnG , > > You can refer to https://github.com/MILVLG/mcan-vqa for the implementaion. We may release the code in the future. > > Best, > > Zhihao @CurryYuan 非常感谢您的伟大的工作，我想请教一下您用的MCAN里面的Co-Attention对三个打分进行加权求和，而且比将三个打分进行逐元素相加效果好，请问您是用的MCAN里面的哪一部分对得到的最后三个打分进行加权求和的呢？是用的Stacking,encoder-decoder，还是attention-reduce？期待您的回复，谢谢！...
> 我不认为你可以简单地做到这一点。STCN 需要多次前向传递，而这些都依赖于不断增长的内存库。一种可能的方法是独立计算键/值编码器/解码器的 FLOP，并根据内存大小手动计算内存读出所需的 FLOP。然后你可以把它们加起来。总 FLOP 将取决于视频长度等。 是的，我们尝试了很多办法和很多包去算STCN 的FLOPs，但是都没有成功，主要感觉STCN模型有多个输入，不像resnet输入只有data.
>  > TODO: vectorise this code!What does this sentence mean?Is the code for reading R, G, and B correct?How do you read R, G, B? My dimension is [batch,channel,H,W]....
> `python learning/evaluate.py --dataset vkitti --odir results/vkitti/best --cvfold 123456` should give you a breakdown of OA and IoU by classes. Thanks.
> `python learning/evaluate.py --dataset vkitti --odir results/vkitti/best --cvfold 123456` should give you a breakdown of OA and IoU by classes. Have you recorded IoU of each category before?Can you just...
Yes, your method has a low argument count, but what about computational complexity? Can the specific FLOPs be given in Partition Part of the Semantic segmentation Part on S3DIS so...
> 你好，打扰您了，我想请问一下，您论文实验结果部分的可视化点云图是怎么操作得到的？ > […](#) > ---Original--- From: "Yongcheng"
can not send email to you!!!
> hi @longmalongma, to visualize either semantic or instance labels, check out this function: > > https://github.com/Yang7879/3D-BoNet/blob/07c99e58f7f1a646c211c150cb7b3c363fdd69d9/helper_data_plot.py#L37 Thanks for your reply,but I find your code can not save results of...
Hi,thanks for your reply.you maybe misunderstanding my meanings.I only want to know how to save your results of segmentation as .ply or .obj files in your code. becase .ply or...