Featured DevTeam.Space Projects
AI, B2B, Backend, ML, Python, Recommendation Engine, TensorFlow
Eksmo is one of the largest publishing houses in Russia. Our machine learning dev team developed neural networks for classifying books by genres. The final system allowed the customer to create a recommendation engine featuring certain books on the publishing house’s advertisements.
DetailsSome of Anton’s Projects
Dozor
For the Dozor Project with a car-sharing company, Anton developed an Android app featuring facial recognition to verify authorized drivers and a novel smoking detection capability, designed from scratch. This required extensive research, data collection, and experimentation with neural networks, resulting in an accurate, offline-capable system for embedded car systems.
For the Dozor Project with a car-sharing company, Anton developed an Android app featuring facial recognition to verify authorized drivers and a novel smoking detection capability, designed from scratch. This required extensive research, data collection, and experimentation with neural networks,...
Read moreBusFactor
In the BusFactor project, Anton focused on monitoring public transport drivers for mobile phone use while driving, ensuring correct camera alignment and continuous operation within the driver's cabin. He successfully managed the deployment of these models onto the client's systems.
Read moreRecognition of goods on the shelves
For a merchandising solution, Anton worked on recognizing goods on shelves, comparing actual product arrangements with planned layouts. This involved OCR, one-shot learning, image generation in Blender, and the development of marking tools for real products. He explored various detection and classification methods, optimized models for Android devices, and experimented with OCR and one-shot learning techniques.
For a merchandising solution, Anton worked on recognizing goods on shelves, comparing actual product arrangements with planned layouts. This involved OCR, one-shot learning, image generation in Blender, and the development of marking tools for real products. He explored various detection and...
Read moreHitfactor
In the Hitfactor project, he contributed to developing a system to detect athletic starting signals or shots in soundtracks using convolutional and recurrent neural networks. Anton improved the system's accuracy in identifying relevant sounds by employing clustering algorithms, distinguishing important shots from background noise.
In the Hitfactor project, he contributed to developing a system to detect athletic starting signals or shots in soundtracks using convolutional and recurrent neural networks. Anton improved the system's accuracy in identifying relevant sounds by employing clustering algorithms, distinguishing...
Read more