Bird Distribution Mapper 9001

• 

  data example 1

• 

  heatmap example 1

• 

  data example 2

• 

  heatmap example 2

Inspiration

We live on a chicken farm want to increase bird quality of life

What it does

It detects chickens using machine learning. Then reports that data in a heat map.

How we built it

After a bit of research we locked into a Faster Regional CNN implementation. Knowing the algorithm we wanted we found an open-source implementation of it on github. We the modified the implementation to work on our computer and our dataset. Once we had a working Neural Net we then created the heatmap portion using seaborn.

Challenges we ran into

The github implementation was broken so we needed to debug and modify it to work on our computers. This involved various edits to the code and the environment. It was very hard to properly display learning rates of both the object detection and classification portions of the Neural Net. We had some documentation and environment issues due to the implementation of the project. Being semi-new to the machine learning scene we had trouble reading and debugging the code.

Accomplishments that we're proud of

it work

What we learned

Use the latest version of tensorflow. Double check your data inputs. Basic machine learning concepts.

What's next for Bird Distribution Mapper 9001

Clean up the repository Create new data over a longer period of time with Raspberry pi's Retrain model for longer period of time

Built With

• conda
• keras
• python
• tensorflow