Semantic Search for IoT-Based Object Recognition Using Clustering and Classification
R. Prasath, V. Vasantha Kumar, V. Jaisankari, R. Siva Subramanian · Advances in computational intelligence and robotics book series · 2025
A search engine helps retrieve information from large datasets using data retrieval techniques. Enterprise search engines analyze diverse organizational data, while open-source engines focus on customization. For code acquisition, data (code, docstrings) is processed into pairs for training models. Optimizations like B-tree-based P2P indexing and syntax-aware tokenization improve efficiency. Real-time object recognition requires at least 10 fps, which may be achievable with newer Intel NCS versions. The system also scales well for IoT applications, achieving 80% accuracy.