Low cost certified welfare of cows

Inspiration

Our inspiration is to develop an end-to-end solution that brings the opportunity to change something about the reality of how we live.

What it does

The main idea of the project is to create an end-to-end system to automate the process of collecting data on the status of animals on a farm to certify the degree of meat quality.

To do this, it is necessary to measure data such as temperature, CO2, dB, and the amount of time the cows remain outside the enclosure. For our project, we have focused on the latter, as it involves greater software development.

Most farms are located in geographical areas where the available bandwidth is very small. For this reason, the system needs an ultra-light data transmission system, and sends as little data as possible to meet the requirement.

How we built it

As for the part of the counting of cows in a specific area, we have decided to count those that are inside the stable. To do this, we used a Yolo neural network. Because it is a proof of concept, we did the detection with people, as access to public datasets was easier. We used Coco's public dataset.

We have developed a Python script that loads the model, makes the inference, and, given an image, calculates the number of people it sees. This number, using the mqtt protocol, sends the data to a local server.

This local server has another script that receives the messages and saves them in a list, in order to create a periodic report for the client on the state of his farm in the future.

Challenges we ran into

During the course of this project we have had to face innumerable challenges, the first of which, and which has been far surpassed, has been to implement a deep learning algorithm in a light device such as a raspberry pi 3b+. It has been a challenge for the team to deploy the communication protocol.

also and no less important we have had to deal with different technical problems, which delayed us several hours in the execution of the project

Accomplishments that we're proud of.

This local server has another script that receives the messages and saves them in a list, in order to create a periodic report for the client on the state of his farm in the future.

What we learned

During these 36 hours of intense work, the team has learned about deep learning model structures and inference, as well as deploying a M2M communication link protocol.

During the course of this work, learning to program the API's of the google cloud services platform has been deepened..

What's next for Low cost certified welfare of cows:

The presented system is a proof of concept, there is still a lot of work to be done, such as implementing a sigfox network in the IoT device, even installing a solar panel system.

The reality is that it is a product that could be commercialized.