search

hongxiaoy / ISO

[ECCV 2024] Monocular Occupancy Prediction for Scalable Indoor Scenes
https://hongxiaoy.github.io/ISO
Apache License 2.0
29 stars 1 forks source link

Visualization for Occ-ScanNet #5

Open chenhaomingbob opened 2 months ago

chenhaomingbob commented 2 months ago

Thanks for your great work! I meet a problem when I visualize the Occ-ScanNet data. Below, I will present the occupancy ground truth and the rgb image (sample is occscannet/gathered_data/scene0002_00/00002.pkl)

occupancy ground truth

image

rgb image

image

According to the visualization, occupancy and image are not aligned. But, I can't find the problem. Can you help me? I use the visualization script of iso/scripts/visualization/NYU_vis_pred.py.

hongxiaoy commented 2 months ago

The visualization script is updated iso/scripts/visualization/OccScanNet_vis_pred.py. Please have a look at it and let me know whether it works properly.

chenhaomingbob commented 2 months ago

Thanks~ I have tested the OccScanNet_vis_pred.py file, and this visualization script works to visualize the occupancy. Here, I paste my visualization results of ISO_mm/data/occscannet/preprocessed_voxels/scene0000_00/00000.npy. It's great. image

However, how can I obtain the sample visualization including occupancy and camera, like Fig.3 in ISO paper? image

hongxiaoy commented 1 month ago

This figure is generated by the matplotlib package. I take the screen shot of the visualization window as a jpg, and create a subplot using plt.subplots function. It's a bit tedious.