Dear organizers, I have some questions regarding the data description of Subtask 2. According to the description on the official website under **iMed Challenge > Data**, the `pose.txt` file is described as follows: > Row 0 corresponds to endoscope1-left and is the identity pose. > Row 1 corresponds to endoscope2-left. However, I noticed that in the provided baseline code, the implementation of the `_parse_poses()` function in the `IMED_Dataset` class seems to use a different convention: Row 0 represents the pose of the training view (`endo2`), while Row 1 represents the pose of the testing view (`endo1`). ``` c2w_by_cam = {} for line in lines: parts = line.split() assert len(parts) == 8, f"Pose row must have 8 values: {line}" cam_id = int(parts[0]) t = np.array([float(v) for v in parts[1:4]], dtype=np.float32) q = np.array([float(v) for v in parts[4:8]], dtype=np.float32) rot = R.from_quat(q).as_matrix().astype(np.float32) c2w = np.eye(4, dtype=np.float32) c2w[:3, :3] = rot c2w[:3, 3] = t c2w_by_cam[cam_id] = c2w assert 0 in c2w_by_cam and 1 in c2w_by_cam, "pose.txt must provide camera ids 0 and 1" return {"cam2": c2w_by_cam[0], "cam1": c2w_by_cam[1]} ``` This appears to be inconsistent with the description on the website. I believe the baseline code itself is correct, and perhaps the description on the website needs to be updated accordingly. If my understanding is incorrect, please kindly correct me. Thank you very much for your clarification. Best regards, Peter