So i am making a object detection model for camera but i have a problem and that is the model size is always to big for the cameras flash memory by a few kb how could i reduce this size. Also if anyone is willing to help and give any tips on how to make the process less of a hassle would be great because currently i am uploading 400 photos to edge impulse and have to manually draw a rectangle around the object on every photo is there a faster way because i plan on adding more photos and it would just take forever.
Hello @destinko497,
-
Reduce the size of your model. If you are working with an object detection model. You can try the FOMO architecture with a 96x96 images. This should be able to run on microcontrollers. If you don’t need the colors, you can also use greyscale images to reduce the size of your model.
-
Auto/assisted labeling.
You can have a look at Labeling queue (Object Detection) | Edge Impulse Documentation and Auto-labeler (Object Detection) | Edge Impulse Documentation
Best,
Louis