I have successfully deployed Edge Impulse using an Arduino Nano BLE sense for audio detection.
I now have a new use case for object recognition, namely, I wish to count traffic and categorise the vehicles into the following groups:
Car
Truck
Bus / Coach
Motorcycle
Does anyone have any comment on using the ESP EYE, as all of the demos online just have one or two object types. Before I get started, I would like to know if this is likely to be too much for Microcontroller such as the ESP?
Object tracking is what you are after here, have a look at our FOMO which was created specifically for constrained devices:
Tutorial
Entertaining real world look at FOMO by Shawn here, here he’s using a more powerful MCU than the ESP32 but still you can get reasonable performance from the esp32 once you can deal with slightly slower counting / inference :
I am sure edge impulse can easily do this object tracking what I am curious is to see if the ESPEYE can handle it
I have previously used the ESP32CAM for object detection on edge impulse. After trying everything I was able to get close to acceptable results with 3 models.
Would love to know how your ESPEYE performed at par!
I have restarted my project and will let you know how it progresses. So far I have created the database of around a 1000 images and achieved a 71% score. Once I have this over 90%, I will build and deploy it to my ESP32-CAM and test.
