Yogesh Yadav

Intrusion Detection System Using YOLO and OpenCV
Implemented a real-time intrusion detection system utilizing the YOLO object detection model and OpenCV for animal detection. Developed a PyQt-based GUI for monitoring, and integrated Flask for backend processing to generate detailed incident reports. The system delivers efficient detection and reporting, improving response times and accuracy.

Military Vehicle Detection with YOLO and MySQL Integration
Deployed the YOLO object detection model to accurately identify and classify various military vehicles from aerial footage. Integrated the system with a MySQL database to log detection data and used Flask for generating comprehensive reports. Optimized for high-speed detection and record-keeping to assist in military surveillance and intelligence operations.

Satellite Object Detection for GIS Analysis Using RCNN and YOLO
Applied RCNN and YOLO models to satellite and aerial images for multi-class object detection. Converted detections into vector data, enabling advanced GIS analysis such as clustering and precise object localization. This solution facilitated spatial analysis, aiding in land use planning and infrastructure optimization through enhanced decision-making capabilities.

Satellite Change Detection Using Semantic Segmentation
Utilized semantic segmentation models to generate detailed masks from satellite images for detecting environmental and infrastructural changes over time. This method enabled the identification of subtle shifts in landscapes and facilitated data-driven decision-making for urban planning and disaster management.

Drone Detection Using Deep Learning Models
Experimented with and implemented various deep learning models to successfully detect flying drones in real-time scenarios. Optimized detection pipelines to handle different environmental conditions, ensuring accurate identification and tracking for aerial surveillance applications.

AI-Driven Grid Route Optimization for Tata Power Using LULC and Satellite Imagery
Developed an AI-powered solution for optimizing transmission grid routes using Land Use Land Cover (LULC) classification and satellite imagery. Assigned pixel-based weights to analyze terrain, enabling efficient route planning for transmission towers while minimizing ecological impact. This system improved cost-efficiency and reduced transmission line distances in large-scale power projects.

Video Transcription Tool with Timestamping, Note-Taking, and Editing
Engineered a sophisticated video transcription tool leveraging advanced speech-to-text models. The system generates time-stamped transcripts and offers a user interface for note-taking and transcription editing. Designed with a Flask backend for easy data handling and editing functionalities, streamlining the process of video documentation and review.

Video Summarization and Q&A System Using Retrieval-Augmented Generation (RAG)
Built a RAG model-based platform that converts uploaded videos into text, generates concise summaries, and allows users to query the content with natural language. The system retrieves relevant text to answer user questions accurately, enhancing video comprehension and providing an interactive experience for content analysis.

German to Indic Language Translation and Text-to-Speech (TTS) System
Developed a multilingual translation model that converts German text to multiple Indic languages. Integrated the Bhashini TTS engine to transform translated text into audio, providing a complete solution for language translation and accessibility. This system supports seamless communication and information dissemination across diverse language groups.

GPU-Optimized Automated Historical Document Processing
Designed and optimized an automated document processing system to efficiently crop, dewarp, deskew, and enhance historical document images. Leveraged GPU acceleration to significantly reduce processing times, enabling large-scale digitization and preservation of archival documents with minimal manual intervention.