Function get_scene_cameras

visualize_benchmark_web.py:1694–1711

Return camera positions and frustum data for all frames. Binary format: int32(N) | float32[N*12](poses 3x4) | float32[9](K 3x3)

(scene_id: str, z_far: float = Query(10.0), depth_mask: bool = Query(True), conf_threshold: float = Query(0.0))

@app.get("/api/scene/{scene_id}/cameras")
def get_scene_cameras(scene_id: str, z_far: float = Query(10.0), depth_mask: bool = Query(True), conf_threshold: float = Query(0.0)):
    """Return camera positions and frustum data for all frames.

    Binary format: int32(N) | float32[N*12](poses 3x4) | float32[9](K 3x3)
    """
    scene = _find_scene(scene_id)
    if scene is None:
        return Response(status_code=404)

    data = _load_scene_data_raw(scene, z_far, use_depth_mask=depth_mask, conf_threshold=conf_threshold)
    K = data["K"]
    poses = np.array(data["extrinsics"], dtype=np.float32)

    buf = io.BytesIO()
    buf.write(struct.pack("<i", len(poses)))
    buf.write(poses.tobytes())
    buf.write(K.astype(np.float32).tobytes())
    return Response(content=buf.getvalue(), media_type="application/octet-stream")


def _find_scene(scene_id: str):