VisualBacter

An interactive platform for uploading images and running client-side inference with TensorFlow.js to classify bacteria associated with foodborne illnesses.

Features

• Convolutional neural network (CNN) model exported for in-browser use
• Dynamic per-class probability chart (ECharts)
• Preprocessing: grayscale conversion and 150x150 resizing
• Responsive UX with custom UI components and TailwindCSS
• List of supported classes with the highest probability highlighted

Tech

Next.js, React, TensorFlow.js, TailwindCSS, ECharts.

Inference flow

1. Image upload (JPG/PNG)
2. Conversion to tensor and normalization
3. Resizing and request to the /api/prediction endpoint
4. Visualization of the distribution and the dominant class

Motivation

To bring computer vision techniques closer to quick exploratory analysis of foodborne pathogens in an educational, demonstrative setting.