CV Stands for Cat Vision

Nova chillin
Sagan and Nova annotated

Inspiration

We initially wanted to build an automated turret to spray water at the cats when they got on the counter, however hardware and copmputer vision are both new to me so we only got halfway there. I settled for a bunch of tools that effectively detects Nova or Sagan, generates a GPT response berading them then uses a Voice clone of Kevin sent to a google home to yell at them to get of the COUNTER

What it does

Currenty we have all the pieces, a CV model to detect Nova and Sagan, Voice cloning of Kevin and Seb, GPT templates for generating funny berating sentences on the fly

How we built it

We used YOLOV8 for the CV model and we used Eleven labs for the voice cloning, the rest of the tools use a rest api to communicate using Flask, but we also have a bunch of Google cloud functions to use. The Model was trained in a google colab using the rented T4 GPU

Challenges we ran into

Neither of us have ever done computer vision or knew what to do or have ever had a hand in hardware, so this project was alot of dreaming and tons of going "How the hell do you do this"

Accomplishments that we're proud of

"Training the model to do live inference from a laptop came and pretty accurately detect Nova and Sagan is something we are very proud of as we learned on the fly what makes good and bad data sets, the ins and outs of putting a model to training on higher end gpus and inferencing at near real time speeds

What we learned

We learned alot, we trained this model and built 3 versions of our dataset, It was also interesting to use Eleven labs to do voice cloning and seeing what speech patterns where best.