Is there any limitations on the ACAP object-detection-cv25 example?

[jaamatos]

I am currently running the object-detection-cv25 example from the acap-native-sdk-examples on my M3085-V camera but have encountered some issues with detection performance. Specifically, the application fails to consistently detect certain objects in the image, such as a clock and a stop sign mounted on my wall.

Here are the steps I’ve taken so far:

• Modified the manifest.json file to adjust parameters.
• Reduced the detection threshold to 25% to improve object recognition sensitivity.

Despite these changes, the stop sign often fails to appear in the detections, even though the accuracy for other objects seems satisfactory.

My questions are as follows:

1. Are there any known limitations with the COCO library or any other files from that specific app (object-detection-cv25) that might explain this behavior?
2. Am I potentially overlooking any critical steps or adjustments, either in the manifest.json file or elsewhere in the application configuration?
3. Are there additional recommended parameters or adjustments I can make to improve detection consistency for objects like the stop sign?

I would appreciate any guidance or suggestions to resolve this issue. Thank you for your support!

Best regards,
Jose

[github-actions[bot]]

This automatically generated reply acts as a friendly reminder.

Answers to your questions will most often come from the community, from developers like yourself. You will, from time to time, find that Axis employees answers some of the questions, but this is not a guarantee. Think of the discussion forum as a complement to other support channels, not a replacement to any of them.

If your question remains unanswered for a period of time, please revisit it to see whether it can be improved by following the guidelines listed in Axis support guidelines.

[vivekatoffice]

Hi @jaamatos ,
In my understanding, tensorflow-to-larod-cv25 is just an example to help you to explain the working so you can use your model with CV25 devices and you may need to benchmark and accept the hardware capabilities.