search

thusiyuan / cooperative_scene_parsing

Code for NeurIPS 2018: Cooperative Holisctic Scene Understanding: Unifying 3D Object, Layout, and Camera Pose Estimation
http://siyuanhuang.com/cooperative_parsing/main.html
MIT License
101 stars 19 forks source link

hard-coded path in the pickle file #2

Closed ZheC closed 5 years ago

ZheC commented 5 years ago

You have some hard-coded path in the pickle files, e.g., in this line: https://github.com/thusiyuan/cooperative_scene_parsing/blob/master/preprocess/sunrgbd/sunrgbd_parser.py#L116

img_info['imgrgb_path'] '/home/siyuan/Documents/Dataset/SUNRGBD_ALL/SUNRGBD/kv2/kinect2data/000002_2014-05-26_14-23-37_260595134347_rgbf000103-resize/image/0000103.jpg'

img_info['imgdepth_path'] '/home/siyuan/Documents/Dataset/SUNRGBD_ALL/SUNRGBD/kv2/kinect2data/000002_2014-05-26_14-23-37_260595134347_rgbf000103-resize/depth/0000103.png'

Could you add your code to generate the pickle files as well? Also, it seems in your preprocessed data, there is no depth images, should I download the original SUNRGBD dataset as well?

thusiyuan commented 5 years ago

Hi,

Thank you for pointing it out.

I manually change the root temporarily as in https://github.com/thusiyuan/cooperative_scene_parsing/commit/472417abc8eb60343d848739b7ff2d5fa73e6ba2

You also need to download the original dataset, as added in README.

We will prepare the code for generating the pickles files, hopefully, will release them in next month, maybe after the ICCV deadline.

Thank you.

ZheC commented 5 years ago

Thank you for fixing that. That's helpful!

One more question: in this line: img_info['seg2d'] = loadmat(img_info['seg2d_path'])['seg_2d'] '/home/siyuan/Documents/cvpr2018/data_temp/segmentation/1.mat'

Where can we download the segmentation files?

thusiyuan commented 5 years ago

This is the segmentation map we generate for another project.

You could comment out this line.

ZheC commented 5 years ago

Could you kindly share these segmentation files? I was intended to implement a function to visualize the 3d point cloud with semantic information and the corresponding 3d bounding box. Having these segmentation maps will be super helpful! Thank you for the useful code and preprocessed data!

thusiyuan commented 5 years ago

https://drive.google.com/file/d/1T-POZBhkh1YW9W0JG2qgVSb_tnvjoHXx/view?usp=sharing

This is the segmentation file we use. You can use it after changing the seg root in the pickle file.

To be noticed, the seg results are not ground truth but inferred from a learned segmentation model. If you want the ground truth, you may need to process the original dataset.

If you want to do visualization, it should be enough.