Instant-NGP is an experimental software by Nvidia to generate so called NeRFs from a set of 2D images. In contrast to photogrammetry, this is not mesh based, and it is fun to play with. Asides from the images, the software needs camera positions. They are stored in a file transforms.json. Example:
Included with Instant-NGP are a bunch of scripts to extract these from output of photogrammetry software, such as colmap2nerf.py for Colmap. There is also a 3rd party script agi2nerf for Agisoft Metashape.
Is there a way to get extract camera positions as above from the output of Zephyr Lite?
Code:
{ "camera_angle_x": 1.25466864999927, "camera_angle_y": 0.900494284520933, "fl_x": 1886.79279245434, "fl_y": 1886.79279245434, "k1": -6.281414521481763e-05, "k2": 0, "k3": 0, "k4": 0, "p1": 0, "p2": 0, "is_fisheye": false, "cx": 1368.0, "cy": 912.0, "w": 2736.0, "h": 1824.0, "aabb_scale": 16, "frames": [ { "file_path": "./images/_DSC0344.png", "sharpness": 1724.0856494602224, "transform_matrix": [ [ -0.00028271292123410903, 0.43251585105782214, 0.90162628539701, -1.1552158600562916 ], [ 0.999988773481783, 0.004386977114543087, -0.0017909054113789376, -5.676470764927527 ], [ -0.004730008858073797, 0.9016156569609916, -0.43251223566585667, 2.089282683784027 ], [ 0.0, 0.0, 0.0, 1.0 ] ] }, { "file_path": "./images/_DSC0354.png", ?[...]
Is there a way to get extract camera positions as above from the output of Zephyr Lite?
Comment