gmchen_sjtu comments

Results 9 comments of


                                            gmchen_sjtu

MatrixXd M issue

Also, when I try to change your code to 2D, I got the following output in the mosek, could you help me about this? MSK_SOL_STA_PRIM_INFEAS_CER [ WARN] [1617519198.368803552, 40.400000000]: In...

2D detector

Hi, I want to know how to combine my 2D-detector's output with your code? Thanks!

Compute the transforms matrix between the object and the camera

Hi, did you save this question?

PA-Loss coding

Hey friend, have you solved this problem yet?

How to Render YCB Video dataset based custom data, if its textureless?

Hi, have you solved this question?

Problem while testing real data

> 你好@MathGaron， > > 首先，感谢您分享您的出色工作。该性能看起来很有希望。我正在尝试在论文中重现结果，但对于真实数据却出现了很大的翻译错误。 > > 我为时钟训练了一个模型。训练和验证损失曲线似乎很好。最终的训练和验证MSE损失都在0.012左右。然后，我从https://github.com/lvsn/deeptracking改编了推理脚本，并用它来测试目录clock_occlusion_0中的图像。按照本文中的评估程序，我得到了23mm的较大平移误差。旋转误差似乎很好，为2.7度。我发现较大的翻译错误主要来自Tz。然后，我检查了Tz的训练MSE损失，发现它是0.044，而Tx和Ty的训练MSE损失仅为0.003。我想知道这是否是正常情况，并有几个问题： > > 1. 您能否分享合理的培训和验证MSE损失？0.012对于培训来说足够好了吗？ > 2. 您还可以获得比Tx和Ty大得多的Tz MSE损失吗？ > 3. deeptrack_net.py内部的网络体系结构似乎与本文中的有所不同。filter_size_1是64，而不是96。是否会导致性能差距？ > 4. 您对导致Tz大错误的可能原因有任何想法吗？ > > 期待您的答复并提前致谢！ Hi, how did you create...

SSD Load Failure! File with SemanticAttributes-provided name `data/scene_datasets/habitat-test-scenes/skokloster-castle.scn` exists but failed to load.

Hi, have you solved it?

gmchen_sjtu