Python Jupyter Interactive Kernel crashed when use cv2.imshow()

[lovelyPuppies]

Jupyter Kernel Crash During cv2.imshow()

• Description: Python Jupyter Interactive Kernel crashed.

📅 Written at: 2024-10-17 20:45:56

When it Occurs

• Happens when using cv2.imshow() after initializing PoseC3D models:

  import torch
  from pathlib import Path
  from pyskl.models import Recognizer3D, build_model
  from mmcv.runner import load_checkpoint
  
  pyskl_path: Path = Path.home() / "repo/intel-edge-academy-6/external/pyskl"
  pyskl_data_path: Path = pyskl_path / "data"
  
  checkpoint_path = (
      Path.home()
      / "repo/intel-edge-academy-6/external/pyskl/work_dirs/posec3d/x3d_shallow_ntu60_xsub/joint/best_top1_acc_epoch_24.pth"
  )
  
  checkpoint = torch.load(str(checkpoint_path), map_location="cuda")
  assert isinstance(checkpoint, dict)
  
  posec3d_model_config = execute_code_and_extract_variables(checkpoint["meta"]["config"])
  
  checkpoint_path = (
      pyskl_path
      / "work_dirs/posec3d/x3d_shallow_ntu60_xsub/joint/best_top1_acc_epoch_24.pth"
  )
  
  posec3d_model: Recognizer3D = build_model(posec3d_model_config["model"])
  
  load_checkpoint(posec3d_model, str(checkpoint_path), map_location="cuda")
  
  posec3d_model = torch.compile(model=posec3d_model)
  
  posec3d_model = posec3d_model.cuda().eval()

Output Behavior

• In Jupyter Interactive Kernel:

  The Kernel crashed while executing code in the current cell or a previous cell.
  Please review the code in the cell(s) to identify a possible cause of the failure.
  Click here for more info.
  View Jupyter log for further details.
  

• If run as an executable:

  <no message>
  

Root Cause

• Using cv2.imshow() within Jupyter can cause kernel crashes due to compatibility or memory management issues between OpenCV and the Jupyter environment.

Workaround

• Use from IPython.display import clear_output, display instead of cv2.imshow() to avoid crashes in Jupyter.

🛍️ Example: OpenVINO Monodepth Notebook Example

[lovelyPuppies]

이 문제는 다음 요인들이 결합되어 발생한 것으로 보임

• VS Code의 Remote-SSH를 통해 Jetson Nano에 접속.
• Jupyter Notebook (IPython) 실행.
• 하드웨어 가속을 사용하지 않는 OpenCV 활용.

➡️ CPU 사용량 증가 및 메모리 부족 이 원인.