Skip to content

Motion Industries Image-to-Product Matching System

Documentation Homepage

Welcome to the documentation site for the Motion Industries Capstone Project.
This project implements an automated system for discovering, ranking, and validating product images for Motion’s 2M+ item catalog. The system combines Python-based scraping, machine learning–driven confidence scoring, ElasticSearch indexing, and a full-stack review UI to streamline Motion’s internal content workflows.

Project Overview

This system automates the process of finding and validating product images by:

  1. Scraping manufacturer and distributor websites for product images
  2. Ranking images using a machine-learning model
  3. Indexing all results into an ElasticSearch cluster
  4. Allowing human reviewers to approve or reject images through a dedicated UI
  5. Finalizing decisions via an admin-only workflow
  6. Storing user accounts and permissions with NextAuth and Postgres/SQLite

It is designed to scale, improve precision, and reduce the time spent manually sourcing product images.
This meets key project requirements such as automated discovery, ML scoring, review/validation UI, scalable indexing, and documentation.

Features Covered in This Documentation

  • How to install and run each component
  • How the UI features work (Login, SKU Workbench, Catalog Navigator, Review History)
  • How the admin workflow operates
  • How the Python scraper integrates with ElasticSearch
  • How to extend or modify the system
  • Frequently asked questions and troubleshooting

Use the navigation sidebar to explore each section.

System Architecture (High-Level)

 Excel Files ──► Python Scraper ──► ML Scoring ──► ElasticSearch Index
                                                        │
                                                        ▼
                                                Next.js Review UI
                                            (Reviewer & Admin workflow)

Core components:
- Python Scraper – gathers candidate images and metadata
- ML Model – assigns confidence scores
- ElasticSearch – stores searchable product records
- Next.js UI – human review interface
- NextAuth + Postgres/SQLite – user authentication and roles

Roles in the System

Reviewer

  • Views images from the scraper
  • Marks them as pending-approve or pending-reject

Admin

  • Finalizes decisions (approved/rejected)
  • Views the User Index
  • Has full access to workflow pipelines

Documentation Sections

  • Installation – How to set up the scraper, UI, databases, and search cluster
  • Features – How each UI page and ML/scraper feature works
  • Modify & Extend – How to build on or contribute to the codebase
  • Demo Video – Screencast of the system running
  • FAQ – Common issues and fixes

Use the sidebar to access each section.

Project Goal

The system must:
- Automatically discover and match product images using available data
- Score image relevance using ML-based confidence ranking
- Provide a user-friendly interface for verifying and approving images
- Support scaling to millions of products
- Achieve high precision (>90%) on reviewed samples
- Offer documentation for future teams and Motion developers

Summary

This documentation accompanies the full Motion Industries Image-to-Product Matching pipeline.
It is intended for:

  • Developers extending the project
  • Reviewers and admins using the UI
  • Future teams inheriting the codebase
  • Technical evaluators assessing system functionality

Use the navigation menu to begin with Installation, then explore each major system feature.