FanglinBao / HADAR

This is an LWIR stereo-hyperspectral database to develop HADAR algorithms for thermal navigation. Based on this database, one can develop algorithms for TeX decomposition to generate TeX vision. One can also develop algorithms about object detection, semantic or scene segmentation, optical or scene flow, stereo depth etc. based on TeX vision instead of traditional RGB or thermal vision.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

How to test my own dataset in the HADAR-network?

keke-2000 opened this issue · comments

Hi, FanglinBao
I am very interested in your work, but when I try to input my own hyperspectral data into the HADAR network for testing, I have a problem. I hope to input only hyperspectral data without other data, like T,e,v grountruth, but your code seems to load all these data while not only training and verification, but also testing, which makes me confused.

I have solved this problem, thanks

commented

Hi, I am wondering whether your own data is captured by a certain hyperspectral imager and how many channels the data contains? Do you directly used the author's TeX-Net python code to test your own data or you made some modifications to the original code, like the dataloader part? How about the results? Have you compared the TeX-Net results with TeX-SGD results?And do you successfully visualize the results(to get the TeX-vision)? I appreciate it if you're willing to answer my questions.

Hi,
1.the testing hyperspectral dataset is generated by simulations, which contains 7 channels.
2.I do some modifications, add 'test_step' function in model.py
3.the result is bad, because my dataset is quite different from the HADAR dataset, like objects in them, so the Tex-Net's perfommance is poor.
4.I haven't done it, because SGD method is so slow, using near 20 hours on my computer.
5.yes

by the way, I graduated from NJU last year, hhh

commented

校友哈哈哈,感谢及时回复

@keke-2000 请问你有联系方式吗?我们可以交流一下。我的邮箱 cfx2010ep@hotmail.com