MIOS Cage Cam

We have updated the 3D design several times and it now looks and works great!

I was able to test the algorithm on a large set of images from 2 different cage setups. The results are mixed but generally positive. Most of the waste pellets are detected by the segmentation algorithm however dark spots, shadows, and the exposed cage floor still prove to be challenges. I am shifting the focus of development to creating methods to help filter out these false positives.

In parallel, we are developing a validation setup that includes a separate higher definition camera to help visually verify the presence of detected waste pellets. This setup will only be used to evaluate the module's accuracy using the different perspective and higher definition camera to more easily view the cage floor. 

Project Created on Hackaday!

Development up to this point included...

1. Initial OpenMV Cam M7 image acquisition and testing
2. Design of 3D Printed Enclosure
3. Initial acquisition in animal facility
4. Troubleshooting of image quality (exposure was too short and captured room light flicker)
5. Added LED control for consistent lighting

Currently, 2 units are deployed in an animal facility to acquire images under varied lighting conditions.