- Import mesh or GS points in blender, and set camera track (see
setCamTrackInBlender.md
) - Run
visEngine/blender/export_cam_track.py
, to export cameras for all frames. These are stored incamera_track_opencv.json
- Read
camera_track_opencv.json
in any project, please refer todemo_track_read.py
install the following packages
- basic
- plyfile: for managing point cloud
- numpy
- skimage
- ace visualization
- torch
- trimesh: for meshing camera traj
- pyrender: for rendering
- Pillow: for PIL Image
- scipy
- matplotlib
- skimage
- bisect
- mono depth
- torch
- kornia: for mono depth to pcd
- Pillow: for PIL Image
- opencv-python: for cv2
- open3d: for point clouds
cameras are stored in camera_info_opencv.json
with following formats:
P_c2w
: camera to world matrix in opencv/colmap system, 4x4K
: camera intrinsic parameter, 3x3width
: image widthheight
: image heightFovX
: fov x, calculated from KFovY
: fov y, calculated from Kimage_path
: relative path of imageimage_name
: image name without .png/.jpg