Update app.py

app.py

@@ -475,10 +475,22 @@ def render_single_frame(
     if not np.isfinite(verts).all():
         blank = np.ones((frame_height, frame_width, 3), dtype=np.uint8) * 200
         return blank
-
+    
+    # Use vertices as-is (rotation should be applied before calling this function)
+    # But if fixed_center is None, we need to compute from unrotated vertices
+    # If fixed_center is provided, it should be from rotated vertices
+    verts_used = verts.copy()
+    
+    # Compute center for camera positioning
+    mesh_center = verts_used.mean(axis=0)
+    if fixed_center is not None:
+        camera_target = fixed_center
+    else:
+        camera_target = mesh_center
+    
     # Create scene
     scene = pyrender.Scene(bg_color=bg_color, ambient_light=[0.4, 0.4, 0.4])
-
+    
     # Material
     material = pyrender.MetallicRoughnessMaterial(
         metallicFactor=0.0,
@@ -486,16 +498,12 @@ def render_single_frame(
         alphaMode='OPAQUE',
         baseColorFactor=color
     )
-
-    # Create mesh
-    mesh = trimesh.Trimesh(vertices=verts, faces=faces)
+    
+    # Create mesh (exactly like compare_vqvae.py)
+    mesh = trimesh.Trimesh(vertices=verts_used, faces=faces)
     mesh_render = pyrender.Mesh.from_trimesh(mesh, material=material, smooth=True)
     scene.add(mesh_render)
 
-    # Compute center for camera positioning
-    mesh_center = verts.mean(axis=0)
-    camera_target = fixed_center if fixed_center is not None else mesh_center
-
     # Camera setup
     camera = pyrender.IntrinsicsCamera(
         fx=focal_length, fy=focal_length,
@@ -503,21 +511,13 @@ def render_single_frame(
         znear=0.1, zfar=20.0
     )
 
-    # Camera pose: After 180-degree rotation around X-axis, coordinate system changes
-    # Camera should be positioned in front (negative Z) with flipped orientation
-    # This matches visualize.py and ensures proper face visibility
+    # Camera pose: positioned BEHIND the subject (at +Z), looking at -Z (toward face)
+    # This matches compare_vqvae.py exactly
     camera_pose = np.eye(4)
     camera_pose[0, 3] = camera_target[0]                    # Center X
-    camera_pose[1, 3] = camera_target[1]                    # Center Y (body center)
-    camera_pose[2, 3] = camera_target[2] - camera_distance  # In front (negative Z)
-
-    # Camera orientation: flip to look at subject (SOKE-style)
-    # This rotation makes camera look toward +Z (at the subject)
-    camera_pose[:3, :3] = np.array([
-        [1,  0,  0],
-        [0, -1,  0],
-        [0,  0, -1]
-    ])
+    camera_pose[1, 3] = camera_target[1]                    # Center Y  
+    camera_pose[2, 3] = camera_target[2] + camera_distance  # BEHIND (positive Z)
+    camera_pose[:3, :3] = np.eye(3)  # Identity = look at -Z
 
     scene.add(camera, pose=camera_pose)
 
@@ -619,12 +619,16 @@ def render_video(
     # Apply rotation to all frames
     # (T, V, 3) dot (3, 3) -> (T, V, 3)
     verts = np.matmul(verts, rot_matrix[:3, :3].T)
+    # Apply orientation fix: rotate 180 degrees around X-axis (like compare_vqvae.py)
+    verts = verts.copy()
+    verts[..., 1:] *= -1
+    
     # Trim last few frames to remove end-of-sequence artifacts
     T_total = verts.shape[0]
     trim_amount = min(8, int(T_total * 0.15))
     T = max(5, T_total - trim_amount)
 
-    # Compute fixed camera target from first frame
+    # Compute fixed camera target from first frame (using rotated vertices)
     fixed_center = verts[0].mean(axis=0)
 
     frames = []
@@ -664,16 +668,11 @@ def render_comparison_video(
     if not ensure_pyrender():
         raise RuntimeError("PyRender not available")
 
-    # Apply orientation fix
+    # Apply orientation fix: rotate 180 degrees around X-axis (like compare_vqvae.py)
     verts1 = verts1.copy()
     verts2 = verts2.copy()
-    # verts1[..., 1:] *= -1
-    # verts2[..., 1:] *= -1
-    rot_matrix = trimesh.transformations.rotation_matrix(np.radians(180), [1, 0, 0])
-    
-    # Apply to both vertex sets
-    verts1 = np.matmul(verts1, rot_matrix[:3, :3].T)
-    verts2 = np.matmul(verts2, rot_matrix[:3, :3].T)
+    verts1[..., 1:] *= -1
+    verts2[..., 1:] *= -1
 
     # Match lengths and trim
     T_total = min(verts1.shape[0], verts2.shape[0])
@@ -925,4 +924,4 @@ if __name__ == "__main__":
         server_name="0.0.0.0",
         server_port=7860,
         share=False
-    )
+    )