Background
Working as a Business Systems Analyst at Nice House of Plastics, I noticed the challenge of quickly identifying chair models by category — standard vs. portable. Rather than manual memorization, I built a machine learning model and served it through a mobile application.
Approach
Training a model required substantial image data. Instead of photographing chairs individually, I used a more efficient workflow:
- ›Record a 2-minute video of each chair model
- ›Import the video into Photoshop and export individual frames as images
- ›Develop a classification model using TensorFlow
- ›Build the mobile interface with Flutter
Result
The application successfully classifies chair types from a phone camera photo. It demonstrated that practical machine learning solutions are achievable without formal ML expertise — the right combination of tools and problem-solving instinct matters more.
Future Direction
Plans include expanding the model to classify all product offerings and potentially creating a recommendation engine for the company website.