Neural network visualization and realtime sensor data
List of parks of a city
Park management: Plants in the park, watering plans, sensors, stats
In our society, cities are developing new ways of smart and eco-friendly management. That is the reason why creating green spaces, such as parks or gardens is becoming a very popular trend. Notwithstanding, the irrigation of these open spaces requires of huge ammounts of water. An smart and efficient irrigation is key to reduce water consumption, one of biggest environmental problems nowadays.
What it does
Ceres uses sensors to gather information about these green areas' environment, such as humidity, temperature or sunlight. In addition, the APIs provided by TM Forum and Salesforce provide key weather measurements. With all this inputs, our software provides watering plans adapted to the current conditions of every green space: rain ratios, humidity, temperature, sunlight... It also takes into account the watering necessities of the different plants existing in the green space. In order to consistently increase the efficiency of our software, we will also develop a neural network that will connect the data gathered by different sensors and increase the efficiency in the watering of these green spaces.
How we built it
Ceres uses sensors, a neural network and a variety of APIs to optimise the watering of these green spaces. Salesforce's and TM Forum's APIs manage information about the different cities and green spaces that use Ceres, as well as the data gathered by the sensors. Ceres also uses deep learning and sensor data to irrigate in the accurate moment and optimise water consumption. We built a Golang server to handle some requests to the neural network which was written in Python using Tensorflow.
Challenges we ran into
First of all, we had to learn how to work with a huge amount of tools we had never worked with. We also had to train a neural network in a very short period of time. We also had to work with a huge variety of different technologies (APIs, sensors, servers, arduinos and neural networks, among others) and integrating them all together was a real challenge.
Accomplishments that we're proud of
Our project has a strong eco-friendly focus. We are also proud of having created a useful product. We worked from a variety of different perspectives (sensors, design, servers, etc.).
What we learned
We learnt how to use technologies we never used before. We have also learnt how to train a neural network within a very short period of time (something we never did before in such circumstances). This project made us have a new point of view over something very easy to overlook, like the irrigation of green spaces. We also learned dynamics of teamwork we never did before.
What's next for Ceres
The neural network will improve its efficiency over the future as more cities use it, because more cities = more data, and more data = more training and better performance. We developed Ceres as a smart irrigation tool for cities, but it could also be used to improve the efficiency in the watering of crops.