zswang666 / stereo-lidar-ccvnorm Goto Github PK

Hi, thx for sharing such a good work.
You show evaluation on the KITTI Depth Completion Dataset in TABLE Ⅱ, would you please tell me you train on the full KITTI Depth Completion Train Dataset or part of the train dataset like TABLE Ⅰ.
thank you !

Hello, I am a student who is interested in sensor fusion.

I couldn't hide my surprise while looking at your project and paper.

I ran this and printed it out as follows, but I confirmed that I was using a CPU, not a GPU.

I wonder why you don't use the GPU even though it's set as follows at options.py.

How can I change it to GPU use?

IROS BEST PAPER!!!

Thank you for sharing the nice work.
I have a question about the evaluation.
In the paper, you train the network with the KITTI_Completition training dataset except for the 142 images that are also included in the KITTI Stereo dataset.

My questions are

1. For the depth evaluation, did you measure the metric using the KITTI Completion validation dataset?

2. For the disparity evaluation, what ground truth did you exploit for the metric computation? The gt from KITTI Stereo Benchmark? or the gt from KITTI Completion Benchmark?

2-1. If you exploit the gt from KITTI Stereo Benchmark for disparity metric computation, how did you employ the raw lidar point-cloud ?? As fas as I know, Some part of the images in KITTI Stereo Benchmark does not have raw point-clouds information.

2-2. If you exploit the gt from KITTI Completion Benchmark for disparity metric computation, did you only use the 142 images that are also included in KITTI Stereo Benchmark??

Thank you.

lib/python3.5/site-packages/tensorboardX/utils.py", line 103, in convert_to_HWC
AssertionError: size of input tensor and input format are different.
tensor shape: (1, 3, 256, 512), input_format: CHW
请问您遇到过这个问题嘛，是怎么样解决的

Could I test it on my own images and Lidar points without training? How should I test it on the stereo images and lidar points without additional information?

Thank you for sharing your great work.
I'd like to conduct an evaluetion before training, given that the GPU resources are very limited for me. But it seems that I failed to find the pre-train model for KITTI dataset in the project. So, can you share it in the project?
Thanks for your kind sharing again!

你好，请问下你们网络使用的pytorch的版本是？