zhangyp15 / OccFormer

[ICCV 2023] OccFormer: Dual-path Transformer for Vision-based 3D Semantic Occupancy Prediction
https://arxiv.org/abs/2304.05316
Apache License 2.0
330 stars 23 forks source link

depth #7

Open SISTMrL opened 1 year ago

SISTMrL commented 1 year ago
1681996633055

Hello, i run the code of visualization gt_depths with image, why it shows this way. I haven't done the depth estimation before, but i saw the depth images are not shown as this.

zhangyp15 commented 1 year ago

The gt_depths are generated by projecting sparse LiDAR points onto multi-view images. Therefore, the visualized depth follows the pattern of LiDAR scanlines and is different from the commonly used dense depth.

shabiouyang commented 1 year ago

同学,你好,有一个问题请教一下,就是你们这个depth是多帧lidar叠加生成的吗?我看dataset文件夹的lidar2depth.py文件中lidar_points = np.fromfile(results['pts_filename'], 好像只用了一帧lidar,但是你们数据集中有sweeps,这里有点不太理解,望解答