Fix a bug: structured3d extrinsic error!!!

HaoyiZhu · HaoyiZhu · commit 1093eb6434b7 · 2024-03-28T23:58:18.000+08:00
diff --git a/ponder/datasets/preprocessing/structured3d/preprocess_structured3d.py b/ponder/datasets/preprocessing/structured3d/preprocess_structured3d.py
@@ -255,10 +255,22 @@ def parse_scene(
                     if parse_rgbd or plugin_rgbd:
                         depth = depth[:, :, 0]
                         extrinsic = np.eye(4)
-                        extrinsic[:3, :3] = np.linalg.inv(
-                            np.array([[0, 0, 1], [0, -1, 0], [1, 0, 0]]) @ cam_r
-                        )
+                        extrinsic[:3, :3] = cam_r
                         extrinsic[:3, 3] = cam_t
+                        extrinsic = np.array(
+                            [[1, 0, 0, 0], [0, 0, 1, 0], [0, 1, 0, 0], [0, 0, 0, 1]]
+                        ) @ np.linalg.inv(
+                            np.array(
+                                [
+                                    [0, 0, 1, 0],
+                                    [0, -1, 0, 0],
+                                    [1, 0, 0, 0],
+                                    [0, 0, 0, 1],
+                                ]
+                            )
+                            @ np.linalg.inv(extrinsic)
+                        )
+
                         semantic_map = (
                             np.ones_like(depth, dtype=np.int64) * ignore_index
                         )
diff --git a/ponder/datasets/structure3d.py b/ponder/datasets/structure3d.py
@@ -115,14 +115,6 @@ def get_data(self, idx):
             if (rgbd_dicts[i]["depth_mask"]).mean() < 0.25:
                 return self.get_data(idx)
 
-        for d in rgbd_dicts:
-            d["extrinsic"] = np.array(
-                [[1, 0, 0, 0], [0, 0, 1, 0], [0, 1, 0, 0], [0, 0, 0, 1]]
-            ) @ np.linalg.inv(
-                np.array([[0, 0, 1, 0], [0, -1, 0, 0], [1, 0, 0, 0], [0, 0, 0, 1]])
-                @ np.linalg.inv(d["extrinsic"])
-            )
-
         data_dict = dict(
             coord=coord,
             normal=normal,
