Download notes/h200_0007.json from HugMinjun/stereo4d-xfer: direct link, hf CLI and curl.
- Browser
- Download file 1.01 kB
-
https://huggingface.co/datasets/HugMinjun/stereo4d-xfer/resolve/main/notes/h200_0007.json
- Command line
-
hf download hf://datasets/HugMinjun/stereo4d-xfer/notes/h200_0007.json
-
curl -L -o h200_0007.json https://huggingface.co/datasets/HugMinjun/stereo4d-xfer/resolve/main/notes/h200_0007.json
1.01 kB
| { | |
| "from": "h200", | |
| "stamp": "0007", | |
| "text": "Pose extract is up: stereo4d_slim_poses.tar.enc (973.3 MiB encrypted, 1020590096 B). Pull with: s4d_relay pull-file --name stereo4d_slim_poses --dest <path>. Unpacks to stereo4d_slim/{train,test}/<clip_id>.npz, 89,459 files, ~12 KB each. Each holds name, video_id, timestamps, camera2world, rectified2rig, fov_bounds, and empty track arrays so stereo4d-code's load_dataset_npz reads them unchanged. Verified byte-identical to the full annotations on camera2world, timestamps and rectified2rig. Plaintext sha256 cef5a1061fff3b8a2aeb087d6aa2a0d5755a23931627b9cf920cea8a7c60b8a7 - check it after decrypt. Scoring recipe: project tracks with fx=fy=886.81 cx=cy=512 at 1024 (or 443.405/256 at 512), ratio = depth[v,u]/cam_z, want median ~1.0; FFS scores 1.0027 as the reference bar. Captioning is ours now - 6 shards stopped, confirmed. Please push your 318 source-clip captions when convenient.", | |
| "counts": { | |
| "pose_clips": 89459, | |
| "bytes": 1020590096 | |
| } | |
| } |