feat: initialize offline multiview pose tracking experiment

Set up pose_tracking_exp as a uv-managed Python package for offline multiview body tracking experiments. This initial commit includes: - the typed package scaffold, CLI entrypoints, and repo-local uv configuration - scene and replay loaders for generic JSON replays and ActualTest parquet inputs - ParaJumping payload conversion and RTMPose-to-body20 normalization - a custom articulated tracker with tentative, active, and lost lifecycle handling - RPT-backed proposal generation, camera convention handling, and multiview reprojection updates - regression tests for normalization, camera conventions, ActualTest ingestion, seeding, and tracker smoke flows - project documentation covering extrinsic formats and the ActualTest calibration caveat
2026-03-26 13:03:56 +08:00
commit f1a2372b3c
27 changed files with 3791 additions and 0 deletions
@@ -0,0 +1,67 @@
+import json
+from pathlib import Path
+
+import numpy as np
+import pytest
+
+pytest.importorskip("rpt")
+
+from pose_tracking_exp.models import CameraFrame, FrameBundle, PoseDetection, TrackerConfig
+from pose_tracking_exp.replay import load_scene_file
+from pose_tracking_exp.tracker import PoseTracker
+
+RPT_ROOT = Path("/home/crosstyan/Code/RapidPoseTriangulation")
+
+
+def test_tracker_promotes_rpt_sample_person():
+    scene = load_scene_file(RPT_ROOT / "data/p1/sample.json")
+    pose_payload = json.loads((RPT_ROOT / "tests/poses_p1.json").read_text(encoding="utf-8"))
+    view_poses = pose_payload["2D"]
+
+    tracker = PoseTracker(
+        scene,
+        TrackerConfig(
+            tentative_min_age=2,
+            tentative_hits_required=2,
+            tentative_promote_score=1.2,
+            proposal_min_score=0.5,
+        ),
+    )
+
+    bundles: list[FrameBundle] = []
+    for bundle_index in range(3):
+        views: list[CameraFrame] = []
+        for camera, detections in zip(scene.cameras, view_poses, strict=True):
+            pose_array = np.asarray(detections, dtype=np.float64)
+            frame_detections: list[PoseDetection] = []
+            for person_pose in pose_array:
+                mins = person_pose[:, :2].min(axis=0)
+                maxs = person_pose[:, :2].max(axis=0)
+                frame_detections.append(
+                    PoseDetection(
+                        bbox=np.asarray([mins[0], mins[1], maxs[0], maxs[1]], dtype=np.float64),
+                        bbox_confidence=1.0,
+                        keypoints=person_pose,
+                    )
+                )
+            views.append(
+                CameraFrame(
+                    camera_name=camera.name,
+                    frame_index=bundle_index,
+                    timestamp_unix_ns=1_000_000_000 + bundle_index * 33_000_000,
+                    detections=tuple(frame_detections),
+                    source_size=(camera.width, camera.height),
+                )
+            )
+        bundles.append(
+            FrameBundle(
+                bundle_index=bundle_index,
+                timestamp_unix_ns=views[0].timestamp_unix_ns,
+                views=tuple(views),
+            )
+        )
+
+    results = tracker.run(bundles)
+    assert any(result.proposals for result in results)
+    assert any(result.active_tracks for result in results[1:])
+    assert len(results[-1].active_tracks) >= 1