Hey I am currently trying to use the tf object detection under the machine learning examples section in the OpenMV IDE. I keep running into an issue where I receive the ValueError as stated in the topic name. I am using the “load_to_fb = True” command in order to load the file onto the frame buffer but now I am receiving this error. I have tried the various models that are located in the CNN network library and I keep receiving this error. Any guidance or help would be greatly appreciated!
Hi, the CNN mobilenet examples are outdated and will be removed shortly in a new IDE release this week. They don’t run anymore because Google removed support for uint8 networks in the TensorFlow lite for microcontroller package.
The FOMO object detection example works. You can train new classification networks using Edge Impulse.
Ok thank you. Just one question, when I am training my model on Edge Impulse should I be using the “FOMO (Faster Objects, More Objects) MobileNetV2 0.35” model? and would I run into a compatibility error by using YoloV5 or another model?
Right now, the cameras can only use FOMO at acceptable speeds. YOLO runs at 0.1 FPS or so. We’ll have new products out soon which will be much faster.
The mobilenet backbone is fine. It’s just that the models we had in the CNN library were not usable. Previously they were.