Sourabh S Email and Phone Number

- Collaborated on enhancing product capabilities through the integration of vision-language foundation models.- Orchestrated end-to-end management of Python package encompassing development, integration, and testing processes.

- Implemented various EMD variants for enhanced analysis and patch search optimization.- Developed and implemented a 3D bounding box overlay tool.- Explored the impact of different patch featurizers on generating feature maps for diverse datasets.- Contributed to projects focusing on few-shot object detection and open-vocabulary classification and detection techniques.

- Conducted a comprehensive literature review on patch search distances.- Implemented Earth Mover’s Distance on Patchwise features for Search.- Validated the implementation through rigorous experiments, ensuring accuracy.- Optimized the distance calculation, exploring alternatives, conducting profiling, and fine-tuning for enhanced performance.

- Configured CARLA simulator on local machine.- Curated datasets: day-night, random, dynamic weather, and object localization.- Explored simulation data feasibility for annotation.- Implemented YOLO object detectors on mimicked datasets.

- Collect real world garbage images and annotate the images with masks.- Implement a novel mobile and cloud application to segregate garbage using SOTA Mask R-CNN algorithm.- Publish a paper on the outcomes of the project.