AI Image Generation with Stable Diffusion: Blender Add-on

The «Stable Diffusion Blender Add-on» is an innovative tool designed to transform the creative process within Blender Specifically tailored for artists and designers, this add-on allows users to render creative and artistic images directly from Blender’s viewports, streamlining their modeling workflow. Utilizing the concept of stable diffusion, the add-on can generate high-quality renders without the typical time-consuming process associated with traditional Blender rendering.

Cleverfarm(IoT, WebGL)

Interactive digital environment where users can engage with various aspects of farming and agricultural practices

Automatic Number Plate Recognition with YOLO book

We created a comprehensive guide that provides detailed explanations, practical examples, and step-by-step tutorials to help readers master YOLO. It teaches how to train the YOLO model to accurately detect and recognize license plates in images and real-time videos, as well as how to build an end-to-end ANPR system with YOLO from data collection to deployment. The book includes source code, hands-on coding experience, and a step-by-step guide with clear explanations and code examples, allowing readers to gain practical skills that can be applied to real-world projects.

Computer VisionDeep LearningOpenCV

Detect & Track & Count People

Using the state-of-the-art pre-trained models to detect/track people and basic geometry to detect precisely who is entering and sorting.

LabelReader: Advanced OCR for Product Labels

The Product Label OCR project aimed to develop a robust Optical Character Recognition (OCR) system tailored specifically for extracting text information from product labels The system was designed to accurately recognize text on various types of labels, including those with different fonts, sizes, colors, and backgrounds.

WebRTC project

Mind is a videoconference platform for businesses. The system supports video calls with 100 participants, simultaneous screen sharing & simultaneous call recording, and multiple rooms bound to 1 account. Mind pays extra attention to security: encrypted calls, 1-time login verification codes and a peephole function — see who’s trying to attend your conference before letting them in. The system is used by major government institutions: Rosatom, Roskosmos, Central Bank of Russia, major Russian telecom providers MTS, Megafon.

Sketch to 3d

Client wanted a concept art of a sketch he made of a house and turn it into 3d. I used stable diffusion, neural network and custom 3d trained model to produce an image.

Plant Disease Classification and Segmentation

Our objective in this project is exploring the state-of-the-art of deep learning for some agricultural tasks. Also, investigated the best deep learning model for classification (the competition task). And developed a model capable of paying attention to the apple leaf and extracting disease bounding boxes. And deployed the classification model (AWS) to be ready for use in production by farmers using their phone. Skills and deliverables.

Document Intelligence: OCR + OpenAI

Implemented YOLOv11 for high-precision detection in diverse document types like PDFs and invoices. Capable of identifying Text, Tables, Columns, Paragraphs, Key-Value Pairs, Figures, and Logos in documents. RAG based CSV/key-value retrieval from OCR extraction. Trained on 20,000 self-annotated images for tailored accuracy and relevance. Attained an impressive 95% Mean Average Precision (mAP) in object detection. Offers significant value in digital archiving, data extraction, and automated document management.

Contact Us