Challenge

At SoftTeco, we constantly strive to keep up with the newest technologies in order to remain competitive and offer our clients top-notch solutions. With this in mind, we ecided to independently grow expertise in Machine Learning and came up with an ML-powered app for calculating the number of bees in the beehive and building forecasts of the bee population growth and its overall well-being. The biggest challenge was to get the right expertise in object recognition, object detection and correct object classification. At the same time, we were learning how to properly collect the data, process it and create alid datasets for the model.

We chose beehive management due to a number of reasons. First, SoftTeco is a member of the UN Global Compact organization and has a strong focus on sustainability and the monitoring of bee populations has a big impact on our environment. Second, we have experienced beekeepers in our team that would undoubtedly benefit from our solution. After defining the project and its goal, the working canvas for SoftTeco team looked as following:

1. Definition of project objectives and goals
2. Selection of right tools
3. Data processing
4. ML model development
5. Implementation of the model
6. Testing and monitoring
7. Development of the app

Solution

BeeTrack is an AI-powered application that uses computer vision to accurately count the number of bees in the hive. As well, the solution can build intuitive and clear charts and graphs so the user can easily track the dynamics of the hive’s growth and its well-being. SoftTeco solved many ML-related issues before developing the application. The team was responsible for every process involved in the project: from defining the product’s architecture to data labeling and neural network training. This helped us significantly grow expertise in Machine Learning and deliver a fully functioning product as the end result.

How It Works (Neural Model Training)

Before starting to work with the data and neural networks, the SoftTeco team designed a robust product architecture. The architecture consists of three parts: an application, a server and a cloud. When building the architecture, we made sure all of its components interact with each other in a frictionless manner and that the system is scalable and suitable for future app growth.

For successful model training, SoftTeco developers worked with the data closely. The team took care of all data-related tasks: data mining (including data search and labeling) and data processing. It also took 2000 epochs to train the neural network and receive accurate results upon testing

SoftTeco created a model on the basis of the YOLOv3 tool for object recognition and trained this model on Google Cloud Platform. After that, we inserted a trained model into PyTorch and set up a Flask web server. At the same time, we developed an API on the same server for the mobile app.

How It Works (Application Development)

SoftTeco developed a user-friendly iOS application for beekeepers to keep track of their bees and monitor the hive well-being based on the data.

One of the biggest advantages of using the SoftTeco application is the amount of time saved for every inspection. The traditional manual inspection can take up to 1 hour for a single beehive. By using the BeeWell app, a beekeeper can perform an inspection of a hive in 10 minutes or less.

This was achieved due to the automation of the photo-taking process. The application is capable of independently identifying the frame and taking the picture with the best focus to optimize the image recognition results.

The team used Swift5 to build the iOS mobile application. As well, Swift5 was used to develop analytics that visualizes the beehive inspection results in the form of clear charts and graphs.

The app allows users to take photos of their beehives, then sends them directly to the server and returns the inspection results. In the near future, the SoftTeco team plans to add extra functionality to the application to make inspections more advanced and comprehensive.

Results:

SoftTeco successfully developed an AI-powered mobile application for the iOS platform to help beekeepers take better care of their beehives. Our team continues to work on the app and optimize it in order to grow own skills and introduce the product to the international market.