Error Encountered When Processing Custom Data

[poolyone1]

Thank you for developing this excellent project! I'm currently trying to run my custom data (my video_L.mp4 and depth.npz) through the pipeline in "all" mode, but I'm encountering an error that prevents successful execution.

 [Open3D WARNING] [KDTreeFlann::SetRawData] Failed due to no data.███████████████████████████▊| 560/561 [08:46<00:00,  8.45it/s]
[Open3D WARNING] [KDTreeFlann::SetRawData] Failed due to no data.
[2026-01-23 11:37:26,630][phantom.processors.hand_processor][ERROR] - ICP registration failed: [Open3D Error] (std::shared_ptr<open3d::pipelines::registration::Feature> open3d::pipelines::registration::ComputeFPFHFeature(const open3d::geometry::PointCloud&, const open3d::geometry::KDTreeSearchParam&)) /root/Open3D/cpp/open3d/pipelines/registration/Feature.cpp:120: Failed because input point cloud has no normal.

100%|██████████████████████████████████████████████████████████████████████████████████████████| 2/2 [22:59<00:00, 689.89s/it]
----------------- ACTION PROCESSOR -----------------                                                                          
100%|███████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:03<00:00,  1.74s/it]
----------------- SMOOTHING PROCESSOR -----------------
  0%|                                                                                                   | 0/2 [00:00<?, ?it/s]/home/admin128/.local/lib/python3.10/site-packages/sklearn/gaussian_process/kernels.py:440: ConvergenceWarning: The optimal value found for dimension 0 of parameter k2__noise_level is close to the specified lower bound 1e-05. Decreasing the bound and calling fit again may find a better value.
  warnings.warn(
/home/admin128/.local/lib/python3.10/site-packages/sklearn/gaussian_process/kernels.py:440: ConvergenceWarning: The optimal value found for dimension 0 of parameter k1__length_scale is close to the specified lower bound 1e-05. Decreasing the bound and calling fit again may find a better value.
  warnings.warn(
100%|███████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:04<00:00,  2.16s/it]
----------------- HAND_INPAINT PROCESSOR -----------------
load pretrained SPyNet...
load checkpoint from http path: https://download.openmmlab.com/mmediting/restorers/basicvsr/spynet_20210409-c6c1bd09.pth
Processing batches: 100%|███████████████████████████████████████████████████████████████████████| 4/4 [00:13<00:00,  3.38s/it]
Processing batches: 100%|█████████████████████████████████████████████████████████████████████| 12/12 [00:46<00:00,  3.90s/it]
100%|███████████████████████████████████████████████████████████████████████████████████████████| 2/2 [01:10<00:00, 35.06s/it]
----------------- ROBOT_INPAINT PROCESSOR -----------------
Error executing job with overrides: ['demo_name=D435i', 'data_root_dir=../data/raw', 'processed_data_root_dir=../data/processed', 'mode=all', 'render=true', 'bimanual_setup=single_arm', 'target_hand=left', 'input_resolution=720', 'depth_for_overlay=true', 'camera_intrinsics=/home/admin128/JunyuFan/phantom/phantom/camera/D435i_intrinsics_238222075705.json', 'camera_extrinsics=/home/admin128/JunyuFan/phantom/phantom/camera/D435i_extrinsics_238222075705.json']
Traceback (most recent call last):
  File "/home/admin128/JunyuFan/phantom/phantom/process_data.py", line 240, in hydra_main
    main(cfg)
  File "/home/admin128/JunyuFan/phantom/phantom/process_data.py", line 224, in main
    process_all_demos(cfg, processor_classes)
  File "/home/admin128/JunyuFan/phantom/phantom/process_data.py", line 115, in process_all_demos
    processor = processor_cls(cfg)
  File "/home/admin128/JunyuFan/phantom/phantom/processors/robotinpaint_processor.py", line 71, in __init__
    self._initialize_robot()
  File "/home/admin128/JunyuFan/phantom/phantom/processors/robotinpaint_processor.py", line 78, in _initialize_robot
    camera_params = self._get_mujoco_camera_params()
  File "/home/admin128/JunyuFan/phantom/phantom/processors/robotinpaint_processor.py", line 682, in _get_mujoco_camera_params
    img_w, img_h = self._get_image_dimensions()
  File "/home/admin128/JunyuFan/phantom/phantom/processors/robotinpaint_processor.py", line 718, in _get_image_dimensions
    return img_w, img_h
UnboundLocalError: local variable 'img_w' referenced before assignment

Before that, there are some other error message:

[2026-01-23 11:00:09,207][phantom.processors.hand_processor][ERROR] - Error processing frame 195: Edge hand11<00:18, 19.80it/s]
[2026-01-23 11:00:09,240][phantom.processors.hand_processor][ERROR] - Error processing frame 196: Edge hand
[2026-01-23 11:00:09,270][phantom.processors.hand_processor][ERROR] - Error processing frame 197: Edge hand
[2026-01-23 11:00:09,301][phantom.processors.hand_processor][ERROR] - Error processing frame 198: Edge hand

The command I use is
python process_data.py demo_name=D435i data_root_dir=../data/raw processed_data_root_dir=../data/processed mode=all render=true bimanual_setup="single_arm" target_hand="left" input_resolution=720 depth_for_overlay=true camera_intrinsics=/home/admin128/JunyuFan/phantom/phantom/camera/D435i_intrinsics_238222075705.json camera_extrinsics=/home/admin128/JunyuFan/phantom/phantom/camera/D435i_extrinsics_238222075705.json

Environment:
OS: [Ubuntu 22.04]
Python version: [Python 3.10.19]
Installed dependencies: [from install.sh]

Please let me know if you need any further information
I wonder how can I repair it? Please let me know if you need any further information!
Thank you for your time and assistance. ^-^

[Junyu-Fan]

There is something not adaptive to other parameters. In line 718 of robotinpaint_processor.py

    def _get_image_dimensions(self) -> Tuple[int, int]:
        """
        Calculate image dimensions based on input resolution configuration.
        
        Returns:
            Tuple of (width, height) in pixels
        """
        # Epic
        if self.input_resolution == 256:
            img_w = 456 
        # Phantom paper
        elif self.input_resolution == 1080:
            img_w = self.input_resolution * 16 // 9
        img_h = self.input_resolution
        return img_w, img_h

The image width calculation in _get_image_dimensions (line 718) contains hardcoded values that may not adapt to other input_resolution parameters. You should revise this logic if using a custom resolution. Or use a default for different resolutions

        else:
            # default 16:9 aspect ratio
            img_w = int(self.input_resolution * 16 / 9)