Hi, I’m currently developing a small robot for RoboCup RescueLine competitions, I want to use an OpenMV to detect some objects i already have a labeled dataset on Roboflow, i was wondering if there are any alternatives to EdgeImpulse to train image detection models, since EdgeImpulse free plan is pretty limitaing given that it only lets me do a few training epochs
Hi, we have roboflow yolov8 support working: https://youtu.be/aRnn2LeAS4c
However, you’ll need the AE3 and N6. So, it will be something to use in a few months once the new hardware is out.
I understood Edge Impulse has recently loosened the limitations on the free tier, e.g. going from max 20 to max 60 minutes for training, and if I remember correctly, possibility to also use GPU + perhaps some other improvements as well.
I’ve also been able to use exported Keras files from Google Teachable Machine, but due to its limitations I’m mainly using Edge Impulse.