xzh_23

Results 10 issues of xzh_23

作者您好,请问您在训练时id_loss的收敛情况如何呢?我在训练时发现它收敛速度很慢。可以参考一下您的训练log记录吗?感谢!!

Hi, my code settings are as shown in the figure. Do you know why all the evaluation results are 0 ? ![image](https://user-images.githubusercontent.com/80045218/211504509-1900d47b-3aaf-4a4e-9707-8a9689a27b0c.png) ![image](https://user-images.githubusercontent.com/80045218/211504529-31295691-8c4c-453c-9b95-c43ec809004a.png)

您好, 我在使用您提供的boxInst_latest_ckpt.pth权重文件运行demo.py时,在加载权重这里报错了,可以向您请教一下这个应该怎么解决吗? ![image](https://user-images.githubusercontent.com/80045218/222599016-effe1e8f-67de-4ef5-9936-47988c434acb.png) 感谢!!

尊敬的作者您好: 特别感谢您在多目标跟踪领域做出的贡献。 想请教一下MOTRv2结构图中Y0和Y1应该如何理解呢?为什么第一帧的anchors与偏移量相加后要全部与第二帧的YOLOX检测结果拼接作为第二帧的anchors呢?这样再对所有的anchors计算偏移然后向后传递不就会不断累积、产生大量重叠的框,其作用是什么? ![image](https://github.com/megvii-research/MOTRv2/assets/80045218/d3d5e2fc-57c0-4ea1-9466-ade2cb6fabe9) 希望得到您的解答!!万分感谢!!

Hi, when predicting the final heatmap and object size, it looks like you are using normal convolution. Why not use sparse convolution here? ![image](https://github.com/zhanggang001/HEDNet/assets/80045218/1e94faa2-e938-4f9f-a7aa-185f49f03d12)

Dear author, Thanks for your great work! I am a little confused about the correct use of `sparse_add`, when I use a network with `SparseInverseConv`. If I want to use...

Dear author, Thanks for your great work! I would like to ask, where is the specific code implementation for this part? ![image](https://github.com/thomasverelst/blockcopy-video-processing-pytorch/assets/80045218/208e1ad0-7ae9-4908-91fc-06dd140a62c6) Looking forward to your reply. Thanks a lot!!

Dear author, When I was testing the performance of the model, I was confused about the efficiency of the 2D sparse operation, so I did the following experiment and found...

您好, 请问如何解决这一问题 ![image](https://github.com/user-attachments/assets/9d54f8be-df93-494c-8fdc-9655db3687f5)

您好, 请问您的**训练策略**是一个iter内依次处理5帧(第2帧只融合1个历史帧特征,第3帧只融合2个历史帧特征,以此类推),同时计算每帧的损失吗?**测试策略**是逐帧迭代,每1帧都是融合4个历史帧特征吗? 不知道我的理解是否正确,期待您的回复!! 感谢!!