Error when using las2fmap.py

[YamasakiRuna]

Hi, thank for fixing the BUGs and it works great with examples.las.
But when I use the las file you prvides in ftp://user1:User1314@121.192.180.185/dataD/dbnet-2018.zip, which is the prepared data( I use dbnet-2018\train\1\points_16384\0.las for test), it comes the error:

Traceback (most recent call last): File "J:/DBNet-master/tools/las2fmap.py", line 290, in <module> main() File "J:/DBNet-master/tools/las2fmap.py", line 274, in main if get_fmap(p): File "J:/DBNet-master/tools/las2fmap.py", line 233, in get_fmap rotate_about_center(fmap, 180, 1.0)) File "J:/DBNet-master/tools/las2fmap.py", line 98, in rotate_about_center return cv2.warpAffine(src, rot_mat, (int(math.ceil(nw)), int(math.ceil(nh))), flags=cv2.INTER_LANCZOS4) cv2.error: OpenCV(3.4.4) C:\projects\opencv-python\opencv\modules\imgproc\src\imgwarp.cpp:2611: error: (-215:Assertion failed) src.cols > 0 && src.rows > 0 in function 'cv::warpAffine'
which means I could not make fmap by myself, could you please help me?

[wangjksjtu]

This is because down-sampled point clouds are not dense enough to be mapped into the 2D plane. In the paper, original point clouds which contain millions of points (like example.las) in each scene are used to generate feature maps directly to maintain enough information.

[YamasakiRuna]

Sorry for late reply!
Got it! Thanks a lot!
By the way, would you like to share the code with using LSTM? I'm a noob of Tensorflow and want to learn how to do like your paper.

[wangjksjtu]

Thanks for the interest! For now, I have other projects to work on so maybe there is no time to add more models and results. There are also many fantastic tutorials of tensorflow (including LSTM ops) on Github, which are good materials to learn and start from :)

By the way, any pull-requests about new models/results are highly welcomed and appreciated!