RoboPolice

Inspiration

Manual inspection of the entire city camera surveillance can be a tedious tasks, I want to automate them.

What it does

How I built it

I built the model using python with pytorch as the training and inference framework for the deep learning models. The model used for instance segmentation is a Hybrid Task Cascade (HTC) Network with ResNet as the backbone.

Challenges I ran into

Accomplishments that I'm proud of

What I learned

What's next for RoboPolice

Adding more of what violations it could detect including some time-series actions.

Built With

• cuda
• dash
• nvidia
• python
• pytorch