Welcome to the #blog4 of IoT Bee Hive monitor. In a previous blog post, we developed a classification model for detecting the presence of bees.
Our project involves training an object detection model that can identify bees, which we will then use to implement a bee counting system using the FOMO method.
Sensor & Block Information
-
Camera module with input images 96 x 96 pixels - An image block to normalize the image data, and reduce the colour depth to grayscale
-
FOMO transfer learning block-based on MobileNetV2 0.35
To gather data for the project, numerous pictures were taken and the bounding boxes enclosing the bees were labelled.
| {gallery}FOMO - ei |
|---|
|
|
|
|
Designing an impulse
Next, we create an impulse by selecting the Image data, processing block, learning block, and output features.- We have used 96x96 images
- Add Image processing block
- Add the learning block generated from the previous steps
- We have 1 output features (Bee)
Using this dataset, we've trained an object recognition model based on FOMO (Faster Objects, More Objects), which achieved the detection of the bee in the image frame. Live classifications from the Nicla Vision.
Demo
The tinyML model was deployed as a fully-contained OpenMV library. On deployment, Arduino pro nicla vision achieved counting the bee found in the image frame.
Now the bees can be counted from a new image frame, run through the FOMO object recognition model, and log the bee count hourly. We will be integrating the bee count with the Arduino MKR in the next blog. Thanks for keeping up!
