Bounding box for object detection models

Hi, We are using the OPENMV board and was wondering if it is possible to return the bounding box information for objects that are detected/trained.

Is it also possible to input a custom training dataset with bounding boxes indicating the objects to be trained?

Thanks
Wietz

Hi @Wietz, we’re currently not supporting object detection in Edge Impulse (but will at some point in the future).

Do you have any time lines for object detection?

Thanks

@Wietz, not ones that we’re committed on :smiley: But I hope somewhere in Q2.

Hi @janjongboom , I saw you launched the Object detection features. I liked your video on this.
Have you done any tests with OPENMV or the Portenta vision boards?

Thanks,
Wietz

Hi @Wietz, not yet, we have done the work to port any necessary ops to EON but now need smaller models first before we can run them on Cortex-M. No ETA yet but we’re working closely with OpenMV and Arduino to get this in.

1 Like

Hello @Wietz ,

We now have a model that supports object detection on OpenMV cams: https://www.edgeimpulse.com/blog/announcing-fomo-faster-objects-more-objects

Regards,

Louis