Video tab: three temporal clips through the video matchers, mp4 outputs

app.py

@@ -61,6 +61,56 @@ def build_examples():
     return rows
 
 
+CLIP_DIR = os.path.join(os.path.dirname(os.path.abspath(__file__)), "clips")
+VIDEO_CHOICES = [engine.NAMES[k] for k in ("sav", "dynamic", "bida")]
+
+
+def _load_clip(name):
+    import glob
+    d = os.path.join(CLIP_DIR, name)
+    L = [np.asarray(__import__("PIL.Image", fromlist=["Image"]).open(p))
+         for p in sorted(glob.glob(f"{d}/left_*.jpg"))]
+    R = [np.asarray(__import__("PIL.Image", fromlist=["Image"]).open(p))
+         for p in sorted(glob.glob(f"{d}/right_*.jpg"))]
+    return L, R
+
+
+def _write_video(frames, path, fps=8):
+    import imageio.v2 as imageio
+    with imageio.get_writer(path, fps=fps, codec="libx264", quality=7) as w:
+        for f in frames:
+            w.append_data(f)
+    return path
+
+
+@_gpu
+def run_video(clip_name, model_name, show_orig):
+    import tempfile
+    fam = KEY[model_name]
+    L, R = _load_clip(clip_name)
+    if not L:
+        raise gr.Error(f"clip {clip_name} not found")
+    tmp = tempfile.mkdtemp()
+    _write_video(L, f"{tmp}/input.mp4")
+    d_ours = engine.predict_clip(fam, "ours", L, R)
+    vmax = max(8.0, float(np.percentile(np.abs(d_ours), 99)))
+    behind = 100.0 * float((d_ours < -0.5).mean())
+    _write_video([engine.colorize(f, vmax) for f in d_ours], f"{tmp}/ours.mp4")
+    md = (f"**{model_name} (corrected)** — {len(L)} frames, range "
+          f"[{d_ours.min():.1f}, {d_ours.max():.1f}] px, {behind:.0f}% behind the "
+          f"screen. Shared color scale ±{vmax:.0f} px.")
+    if show_orig:
+        d_zs = engine.predict_clip(fam, "zs", L, R)
+        _write_video([engine.colorize(f, vmax) for f in d_zs], f"{tmp}/zs.mp4")
+        zs = gr.update(value=f"{tmp}/zs.mp4", visible=True,
+                       label=f"{model_name} zero-shot (released)")
+        md += (f"\n\n**Zero-shot** — range [{d_zs.min():.1f}, {d_zs.max():.1f}] px.")
+    else:
+        zs = gr.update(value=None, visible=False)
+    return f"{tmp}/input.mp4", gr.update(value=f"{tmp}/ours.mp4",
+                                         label=f"{model_name} + ours — {behind:.0f}% behind"), zs, md
+
+
 with gr.Blocks(title="Unmuzzling Stereo") as demo:
     gr.Markdown(
         "# Unmuzzling Stereo — signed disparity for content behind the screen\n"
@@ -72,21 +122,36 @@ with gr.Blocks(title="Unmuzzling Stereo") as demo:
         "models — fine-tuned on horizontally translated SceneFlow only — recover "
         "it with no architectural change. Red = in front of the screen plane, "
         "blue = behind it.")
-    with gr.Row():
+    with gr.Tab("Image matchers"):
+      with gr.Row():
         left = gr.Image(label="Left view", type="numpy")
         right = gr.Image(label="Right view", type="numpy")
-    with gr.Row():
+      with gr.Row():
         model = gr.Dropdown(CHOICES, value=CHOICES[0], label="Backbone")
         show_orig = gr.Checkbox(value=True, label="Also run the released (zero-shot) model")
         btn = gr.Button("Run", variant="primary")
-    with gr.Row():
+      with gr.Row():
         out_ours = gr.Image(label="Corrected (ours)", interactive=False)
         out_zs = gr.Image(label="Zero-shot (released)", interactive=False)
-    stats = gr.Markdown()
-    btn.click(run, [left, right, model, show_orig], [out_ours, out_zs, stats])
-    ex = build_examples()
-    if ex:
+      stats = gr.Markdown()
+      btn.click(run, [left, right, model, show_orig], [out_ours, out_zs, stats])
+      ex = build_examples()
+      if ex:
         gr.Examples(ex, [left, right, model, show_orig], label="Try a rendered pair (CC-BY Blender open movies)")
 
+    with gr.Tab("Video matchers"):
+      clips = sorted(os.listdir(CLIP_DIR)) if os.path.isdir(CLIP_DIR) else []
+      with gr.Row():
+        clip_sel = gr.Dropdown(clips, value=clips[0] if clips else None, label="Clip")
+        vmodel = gr.Dropdown(VIDEO_CHOICES, value=VIDEO_CHOICES[0], label="Video backbone")
+        vshow = gr.Checkbox(value=True, label="Also run the released (zero-shot) model")
+        vbtn = gr.Button("Run clip", variant="primary")
+      with gr.Row():
+        v_in = gr.Video(label="Input (left view)", interactive=False)
+        v_ours = gr.Video(label="Corrected (ours)", interactive=False)
+        v_zs = gr.Video(label="Zero-shot (released)", interactive=False)
+      vstats = gr.Markdown()
+      vbtn.click(run_video, [clip_sel, vmodel, vshow], [v_in, v_ours, v_zs, vstats])
+
 if __name__ == "__main__":
     demo.launch()