Work Accomplished:
This week, I focused on the initial research and setup phase of the computer vision pipeline for SmartCart.
- Started curating a dataset by sourcing existing grocery product images from GitHub repositories and other publicly available datasets.
- Investigated USFDA food data to map size-to-weight ratios for grocery items, which will be useful for estimating weights when barcode or packaging data is unavailable.
- Reviewed OCR-based text extraction methods for distinguishing visually similar products by identifying brand names and labels.
Progress Status:
I have laid the groundwork for the data acquisition and preprocessing steps necessary for real-time product recognition. The insights gained this week will guide dataset preparation and the initial implementation of YOLOv8.
Next Week’s Goal:
For next week, I plan to move into dataset preparation and initial model setup by:
- Finalizing a structured dataset of Aldi grocery products by combining GitHub-sourced images, USFDA data, and web-scraped product images.
- Implementing OpenCV preprocessing techniques for image enhancement and noise reduction.
- Setting up a baseline YOLOv8 model with an initial subset of grocery product images.
- Developing a basic pipeline to match detected objects with database records, using text recognition and visual similarity scoring.