Finalists in the SPbU Start-up − 2023 contest on neural networks for mining industry and barbershops

The Mineral Technologies and Machine Vision Systems Laboratory project is being developed by a team of students and early-career researchers from St Petersburg. They have created an application that enables mineral identification and automated analysis of bulk mineral mixtures. The algorithm is based on a neural network that helps to identify minerals and their optical properties using computer vision.

Grigorii Bogdanov, Captain of the team and a student at St Petersburg University, said that their development facilitates report generation by identifying mineral grains in a video stream. Computer vision coupled with a machine learning algorithm enable monitoring, analysing, classifying and prospecting minerals.

The application is based on a convolutional neural network, which is most commonly applied to analyse visual imagery. It analyses the visual input, taking into account displacement, scaling, perspective and other types of optical distortions. For mineral identification, artificial intelligence trains a model; that is, the model "learns" how to identify minerals using a library of reference data. At present, the project database contains over 10,000 images of around 50 different minerals from deposits located not only in the Russian Federation but also in South America, Africa and Europe. For each image, an annotation file with coordinates and material description is generated for more accurate and comprehensive analysis. Aleksandr Chumakov, a member of the team, is responsible for compiling the database, collecting the materials and taking photographs. He said that the project collection is unique, because the photos are made taking into consideration the specific features of the convolutional neural network. Thus, the collection is adapted to the project objectives for computer vision training.

The team of the start-up project "Mineral Technologies and Machine Vision Systems Laboratory" is composed of students from St Petersburg University and St Petersburg Mining University and researchers from the Centre for New Technologies "Instruments". The project leader is Grigorii Bogdanov, a student at St Petersburg University. He is responsible for: the geological data for the machine learning algorithms; business plan development; financing; and the product presentation. Product management, marketing and customer relations are handled by Oleg Alikin, Director for Research and Technology at the Centre for New Technologies "Instruments". Aleksandr Chumakov, a doctoral student at St Petersburg University, is compiling the reference database. Aleksei Boikov, a student at St Petersburg Mining University, focuses on the learning algorithms and data processing.

‘Participating in the SPbU Start-up − 2023 contest has been an opportunity to present our project to a large number of people, including experts in geology. The contest helped our team to structure our work in terms of business model alignment. There were very interesting lectures; and we have gained valuable experience and unique insights,’ shared Grigorii Bogdanov, Captain of the team and a student at St Petersburg University.

At the moment, the start-up team "Mineral Technologies and Machine Vision Systems Laboratory" is working with mineral bulk mixtures. The application under development can also be used on: thin sections; polished sections; cores; and other types of rock and mineral materials.

The project initiators maintain that the developed system will accelerate and facilitate the work of mining and mineral processing companies that need to constantly monitor the ore quality and evaluate the important components in the ore. The project has potential for future growth in the materials recycling sector and geological prospecting. The members of the team are confident that their development could speed up mineral identification by more than tenfold.

The contest of innovative, interdisciplinary projects "SPbU Start-up − 2023" is aimed at identifying, stimulating and supporting the development of the most promising projects in view of establishing original, knowledge-intensive and commercially viable start-ups. St Petersburg University provides participants with ample opportunities for professional development and individual growth.

The final round of the contest will be held at St Petersburg University on 25 May 2023, when the finalist teams will present their projects to an expert jury. The winning teams will receive grants to establish small innovative enterprises with the participation of St Petersburg University to implement their innovative projects. In particular, the grant fund includes training of the finalist teams in the non-degree programme "Start-up Development Management" at the Graduate School of Management at St Petersburg University (GSOM SPbU). The programme focuses on: developing market research skills; marketing promotion of products or services; risk management; and other competencies necessary for entrepreneurs. The winners in the SPbU Start-up − 2023 contest who have established small innovative enterprises jointly with St Petersburg University will receive support from St Petersburg University’s Business Incubator.

Another finalist team in the SPbU Start-up − 2023 contest has also used machine learning methods in their project, but in a completely different area. The students from St Petersburg University have developed an application "IT-Barbershop", which will help users to find the right hairstyle for themselves based on their photo and help hair stylists to more effectively promote their services.

The app based on artificial intelligence is able to select a suitable hairstyle for an individual customer from the project’s reference database. Using the customer’s photo, the artificial intelligence algorithm generates a proposal based on: their facial features; head shape; bone structure; length, thickness and colour of their hair; as well as their requests.

The authors of the IT-Barbershop project said that, according to their observations, customers of barbershops often do not know which hairstyles may suit them; hence, they tend to stick to one style. Barbers, on the other hand, do not want to risk delivering inappropriate service and experimenting with a customer’s look. The service-oriented beauty app created by a team from St Petersburg University is designed to help customers to find a unique hairstyle that suits them, and also to find the best stylist for this purpose. The IT-Barbershop project has already compiled a database of barbers. This will enable users to find the right stylist for them, using filters, such as: services provided; price range; and the end-result photos, etc.

‘The software algorithm compares the user’s photo to photos of other people with similar features and chooses an optimal hairstyle for them. In other words, even before going to the barbershop, the app user can choose a hairstyle that will amplify their strengths and hide their weaknesses, while the barber will know in advance what look the customer would like to achieve. The customer just has to show their photo generated by the software to their stylist,’ explained Arman Luarsabian, leader of the IT-Barbershop project and student at St Petersburg University.

The IT-Barbershop start-up team unites students from St Petersburg University: Arman Luarsabian, Captain of the team; Veronika Riabinina, manager and marketer; Viacheslav Shendrikov, lawyer; Aleksei Romanov, chief programmer; and Aleksei Podiachev, business analyst.

The developed app also has a function of online booking system for customers to select the right barber and book their services based on: the customer’s personal needs; goals; salon atmosphere; the barber’s skills; and other parameters. The app includes a customer request form and a map of barbershops. According to the members of the team, the database and list of barbershops will be updated on a regular basis to ensure that the app works efficiently.