Intel Realsense problem

[anbarsanti]

Hi, I was using intel real sense camera D435I and integrated it with YOLOv11.

After some time, my code no longer works.

Here is the code:

# Initialize RealSense Pipeline
pipe = rs.pipeline()
cfg  = rs.config()

# Enable color stream (and depth if you want)
cfg.enable_stream(rs.stream.color, 640,480, rs.format.bgr8, 30)
# cfg.enable_stream(rs.stream.depth, 640,480, rs.format.z16, 30)

# Start Streaming
pipe.start(cfg)

while True:
    frame = pipe.wait_for_frames()
    color_frame = frame.get_color_frame()

    # Convert images to numpy arrays
    color_image = np.asanyarray(color_frame.get_data())

    # Run YOLOv8 OBB inference on the frame
    results = model.track(frame, stream=True, show=True, persist=True, tracker='bytetrack.yaml')  # Tracking with byteTrack

    # Process and visualize the results of Object Tracking with YOLO
    for r in results:
        annotated_frame = r.plot()

        # Display the annotated frame
        cv2.imshow("YOLOv11 OBB Inference - Realsense Camera", annotated_frame)

    # Break the loop if 'q' is pressed
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

# Stop Streaming
pipe.stop()
cv2.destroyAllWindows()

And here is the error message:
File "/home/anbarsanti/Dropbox/YOLOv11/track.py", line 185, in
for r in results:
File "/home/anbarsanti/anaconda3/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 36, in generator_context
response = gen.send(None)
^^^^^^^^^^^^^^
File "/home/anbarsanti/anaconda3/lib/python3.11/site-packages/ultralytics/engine/predictor.py", line 227, in stream_inference
self.setup_source(source if source is not None else self.args.source)
File "/home/anbarsanti/anaconda3/lib/python3.11/site-packages/ultralytics/engine/predictor.py", line 199, in setup_source
self.dataset = load_inference_source(
^^^^^^^^^^^^^^^^^^^^^^
File "/home/anbarsanti/anaconda3/lib/python3.11/site-packages/ultralytics/data/build.py", line 187, in load_inference_source
source, stream, screenshot, from_img, in_memory, tensor = check_source(source)
^^^^^^^^^^^^^^^^^^^^
File "/home/anbarsanti/anaconda3/lib/python3.11/site-packages/ultralytics/data/build.py", line 169, in check_source
raise TypeError("Unsupported image type. For supported types see https://docs.ultralytics.com/modes/predict")
TypeError: Unsupported image type. For supported types see https://docs.ultralytics.com/modes/predict

Could you help?

Thank you

[UltralyticsAssistant]

👋 Hello @anbarsanti, thank you for sharing your issue and for your interest in using Ultralytics 🚀! Below are some pointers and resources that might assist you while an Ultralytics engineer takes a closer look at your report.

From your description, it seems like you're encountering a 🐛 Bug. When reporting an issue, kindly ensure you're providing a minimum reproducible example (MRE). Your shared code snippet is helpful; however, double-check if additional context (such as your YOLOv11 model setup or the input structure) or modifications you’ve made to the pipeline could be relevant as well.

Recommendations

1. Upgrade

   Ensure you are using the latest version of the ultralytics package. This helps verify that the issue isn't already resolved in the latest release. You can upgrade using the following command:

   pip install -U ultralytics

2. Documentation

   The error traceback suggests an issue with the image type being passed as the source. Please review the documentation on supported input sources to confirm your setup is aligned with the expected formats.

3. Environments

   YOLO models, including enhanced versions such as YOLOv11, run seamlessly in various environments, ensuring compatibility across platforms. If your environment might be a factor, consider testing it in one of the following verified environments where all dependencies are preconfigured:

   • Notebooks with free GPU:
     
     
     
   • Cloud VM: GCP | AWS
   • Docker Image: Use Docker Quickstart Guide

4. Debugging the Data Pipeline

   • Review the frame object that's being passed to the model.track() function. Ensure that it's in a valid image or video format compatible with YOLO.
   • Convert your color_image to an acceptable format before inference if necessary, e.g., saving or structuring the numpy array as an image or frame that YOLO recognizes.

Community Support

📢 In the meantime, feel free to engage with the Ultralytics community for real-time support or discussions:

• Join us on Discord 🎧 for chat-based assistance.
• Participate in discussions on our Discourse platform.
• Share your insights or questions on our Subreddit to connect with the broader community.

Status and CI Verification

You can also confirm that all tests are currently passing by checking the Ultralytics CI badge below:

An Ultralytics engineer will assist you soon to help resolve your issue. Thank you for your patience and for being part of the Ultralytics community! 🚀✨

[glenn-jocher]

@anbarsanti the error indicates that the frame object passed to model.track() is not in a format supported by YOLOv11. Instead of passing frame directly, convert color_image (which is the actual numpy array) for inference. Update the model.track() call to use color_image. Let me know if you need further clarification! For supported formats, refer to YOLO Predict Modes.