Presented a research poster titled "Evaluating Body Pose Grouping via Clustering Algorithms" at the School of Physics, Engineering and Computer Science (SPECS) Conference 2026.
The presentation focused on solving the data annotation bottleneck in computer vision, showcasing how unsupervised clustering algorithms (K-Means, DBSCAN, and Agglomerative) can automatically group 3D human movement coordinates into discrete, identifiable poses, using Indian Classical Dance as a controlled ground-truth dataset.
Key discussion points included the custom "Pose Hallucination" data filter and the density-based noise isolation properties of DBSCAN.