Retail & Smart Café Vision Systems
Applied computer vision, stock analysis, and dataset development.
Operational context
My work at Inzpire Solutions applied machine learning and computer vision to retail and café operations, primarily for Aaladipattiyan Karupatti Coffee. It covered stock handling, supply verification, utensil cleanliness, and inventory monitoring.
These were distinct application and data tasks. A stock-transfer workflow and a visual cleanliness model address different operational questions, even within the same business.
My contribution
I worked on stock analysis and verification of supplies transferred from a central warehouse to outlets. On the vision side, I developed classification and detection workflows for clean versus dirty utensils and glasses, along with inventory-monitoring and application-integration work.
Dataset development included image acquisition, annotation, preprocessing, augmentation, splitting, and quality review. Variable lighting, backgrounds, and camera angles were practical challenges for collecting useful examples and applying models consistently.
Data and implementation
Classification assigns a label to an image or crop; detection also locates objects in a scene. The clean/dirty glass work included both forms of dataset development.
Related dataset contributions were Clean/Dirty Glass Classification, Clean/Dirty Glass Object Detection, and Rangoli Detection. Rangoli Detection was an additional contribution; it is not presented as part of the café cleanliness workflow.
Evidence and limitations
An internship report and completion certificate are available below. This account covers dataset and model development and inference/application integration. No sales impact, stock-saving figure, accuracy percentage, or deployment scale is asserted.