fix(review): open a folder of Batch Pose videos as a cohort

CHANGELOG.md

@@ -373,6 +373,12 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- **A folder of Batch Pose videos opens in Session Review.** Videos tracked by
+  Batch Pose (a video beside its `DLC_` CSV) are now sessions of a cohort, as
+  pup-tracked videos already were, and Open recording folder on a folder that
+  is no single recording opens every session in it instead of reporting "no
+  GLIDER recording". A video that already has a session (its ethogram, its pup
+  tracking, or the recording that filmed it) is not listed twice.
 - **Batch Pose follows the animal, not an object the model also detects.** A
   rat model over a novel-object arena also scored each dish as a small rat, and
   the single-animal path kept whichever detection was most confident. The track

src/glider/analysis/cohort.py

@@ -3,7 +3,8 @@
 Session Review opens a folder as a cohort. A cohort mixes three kinds of session:
 apply-run ethograms (``<output>/<video stem>/ethogram_raw.csv``), live
 recordings (a folder of GLIDER tracking/events/data CSVs) and videos, found
-by the ``_pups.npz`` the pup pass leaves beside them. An ethogram usually
+by the ``_pups.npz`` the pup pass or the ``DLC_`` pose CSV the batch pose
+pass leaves beside them. An ethogram usually
 sits beneath the recording it was scored from, and a pup video inside one,
 so each *claims* that recording, and only unclaimed recordings become
 sessions of their own -- otherwise one animal would appear twice in every
@@ -179,9 +180,44 @@ def discover_sessions(root: Path) -> tuple[list[SessionSource], str | None]:
         pup_videos.append(SessionSource(video, recording, session_id_for(recording or video)))
     # Only folders that hold a CSV are classified, so a tree of raw videos is
     # not header-sniffed file by file.
+    recordings: set[Path] = set()
     for folder in sorted({p.parent.resolve() for p in root.rglob("*.csv")}):
-        if folder not in claimed and is_recording(folder):
-            sources.append(SessionSource(folder, folder, session_id_for(folder)))
+        if is_recording(folder):
+            recordings.add(folder)
+            if folder not in claimed:
+                sources.append(SessionSource(folder, folder, session_id_for(folder)))
     sources += pup_videos
+    sources += _pose_videos(root, sources, recordings)
     groups, warning = project_groups(root)
     return [replace(s, group=groups.get(s.session_id, "")) for s in sources], warning
+
+
+def _pose_videos(
+    root: Path, found: list[SessionSource], recordings: set[Path]
+) -> list[SessionSource]:
+    """Videos the batch pose pass tracked that no other session already shows.
+
+    A folder of videos run through Batch Pose holds nothing else -- no
+    ethogram, no recording -- so without these it opened as an empty cohort.
+    A video that already has a session is left to it: an ethogram scored from
+    it, a pup session, or the GLIDER recording that filmed it, whose tracking
+    the recording session reads.
+    """
+    from glider.vision.pose.batch import find_pose_csv
+
+    shown = {s.path.resolve() for s in found if s.path.is_file()}
+    for s in found:
+        if s.path.name == "ethogram_raw.csv" and (video := find_session_video(s.path)):
+            shown.add(Path(video).resolve())
+    sources = []
+    for folder in sorted({p.parent for p in root.rglob("*DLC_*.csv")}):
+        filmed_here = folder.resolve() in recordings
+        for video in sorted(folder.iterdir()):
+            if not is_video(video) or video.resolve() in shown:
+                continue
+            if filmed_here and recording_filmed(folder, video):
+                continue
+            if find_pose_csv(video) is not None:
+                shown.add(video.resolve())
+                sources.append(SessionSource(video.resolve(), None, session_id_for(video)))
+    return sources

src/glider/gui/behavior/analysis_window.py

@@ -897,6 +897,16 @@ def open_path(self, path: Path) -> None:
         if len(ethograms) == 1:
             self.load(ethograms[0])
             return
+        # A folder of sessions -- a batch of tracked videos, several ethograms
+        # -- is what "open this folder" means: all of them, as a cohort.
+        if problem is None:
+            try:
+                sessions = discover_sessions(path)[0]
+            except (OSError, ValueError):
+                sessions = []
+            if sessions:
+                self.load_folder(path)
+                return
         found = "no ethogram_raw.csv" if not ethograms else f"{len(ethograms)} ethograms"
         more = "  Use Open folder as cohort to load them all." if ethograms else ""
         recording = (
@@ -972,8 +982,8 @@ def load_folder(self, root: Path) -> None:
             QMessageBox.warning(
                 self,
                 "Open cohort",
-                f"No sessions found under {root}.\n\nLooked for ethogram_raw.csv files "
-                "and folders of GLIDER recording CSVs.",
+                f"No sessions found under {root}.\n\nLooked for ethogram_raw.csv files, "
+                "folders of GLIDER recording CSVs, and videos with pose or pup tracking.",
             )
             return
         loaded, ids, groups, failed = [], [], [], []

tests/unit/analysis/test_cohort.py

@@ -133,6 +133,37 @@ def test_a_video_with_pups_is_a_session_named_by_its_stem(tmp_path):
     ]
 
 
+def test_a_video_tracked_by_batch_pose_is_a_session(tmp_path):
+    """A folder of videos run through Batch Pose holds nothing else to find."""
+    folder = tmp_path / "rats"
+    folder.mkdir()
+    for stem in ("JP01_OF1", "JP02_OF2"):
+        (folder / f"{stem}.avi").write_bytes(b"")
+        for name in (f"{stem}DLC_rat-1.csv", f"{stem}DLC_rat-1_raw.csv"):
+            (folder / name).write_text("scorer\n", encoding="utf-8")
+    (folder / "untracked.avi").write_bytes(b"")
+    (folder / "JP01_OF1_zones").mkdir()
+    (folder / "JP01_OF1_zones" / "zone_occupancy.csv").write_text("zone\n", encoding="utf-8")
+    sources, _ = discover_sessions(tmp_path)
+    assert [(s.path.name, s.recording, s.session_id) for s in sources] == [
+        ("JP01_OF1.avi", None, "JP01_OF1"),
+        ("JP02_OF2.avi", None, "JP02_OF2"),
+    ]
+
+
+def test_a_pose_tracked_video_with_another_session_is_not_listed_twice(tmp_path):
+    """Its ethogram, its pups, or the recording that filmed it already show it."""
+    recording = _tracking(tmp_path / "rec", video_stem="mouse01")
+    (recording / "mouse01DLC_m.csv").write_text("scorer\n", encoding="utf-8")
+    pups = tmp_path / "pups"
+    pups.mkdir()
+    (pups / "litter.avi").write_bytes(b"")
+    (pups / "litter_pups.npz").write_bytes(b"")
+    (pups / "litterDLC_m.csv").write_text("scorer\n", encoding="utf-8")
+    sources, _ = discover_sessions(tmp_path)
+    assert sorted(s.path.name for s in sources) == ["litter.avi", "rec"]
+
+
 def test_session_id_of_a_video_is_its_stem(tmp_path):
     video = tmp_path / "litter1_pre.avi"
     video.write_bytes(b"")

tests/unit/gui/behavior/test_analysis_window.py

@@ -2851,6 +2851,17 @@ def test_open_video_loads_a_bare_video(self, qtbot, tmp_path, monkeypatch):
         assert len(win._cohort) == 1
         assert win._cohort[0][1].pups is not None
 
+    def test_a_folder_of_pose_tracked_videos_opens_them_all(self, qtbot, tmp_path):
+        """Batch Pose output: videos and their DLC CSVs, no recording, no ethogram."""
+        for stem in ("rat01", "rat02"):
+            _scorable(tmp_path / "batch", stem)
+            (tmp_path / "batch" / f"{stem}_pups.npz").unlink()
+        win = AnalysisWindow()
+        qtbot.addWidget(win)
+        win.open_path(tmp_path / "batch")
+        assert win._ids == ["rat01", "rat02"]
+        assert all(view.pose_path is not None for _, view in win._cohort)
+
     def test_open_path_takes_a_video_file(self, qtbot, tmp_path):
         win = AnalysisWindow()
         qtbot.addWidget(win)