{
  "task": "seg",
  "checked_at": "2026-09-29T06:44:41.711585+00:00",
  "model": "Sapiens2-0.4B",
  "repo": "https://github.com/facebookresearch/sapiens2",
  "commit": "7e5bae88456ac418ff0e58e74106c9fe192055d4",
  "input": {
    "frames": 90,
    "width": 854,
    "height": 480,
    "fps": 24.0,
    "sha256": "1e4fee0782bd9aaf30795c30b05d07806894826d9df38f4cb2af921b3f917de1"
  },
  "frames_processed": 90,
  "elapsed_seconds": 7.217198200058192,
  "model_init_seconds": 9.013367699924856,
  "fps": 12.470213163783466,
  "device": "NVIDIA GeForce RTX 4090",
  "torch": "2.13.0+cu130",
  "precision": "float32 model with CUDA bfloat16 autocast; high float32 matmul precision",
  "peak_allocated_bytes": 2053725696,
  "peak_reserved_bytes": 2493513728,
  "timing_scope": "Single run; excludes checkpoint download and model initialization. Includes decode, detection for pose, preprocessing, inference, rendering and streaming H264 writes; encoder finalization excluded.",
  "weight_sha256": "b85fdb50b7d6123a967d5ee4a505e222baff8d2f7ad6bbf353578c1a61dfbac9",
  "target_selection": "Full-frame segmentation with official preprocessing and 29-class argmax; no temporal smoothing.",
  "accuracy": "No ground-truth annotations used. Confidence is not measured accuracy.",
  "license": "Sapiens2 License; https://github.com/facebookresearch/sapiens2/blob/main/LICENSE.md"
}