Brief
Oral cancer screening web app with image classification and confidence scores.
Ampire is an AI-powered oral cancer screening web application that analyzes inner-mouth images and classifies them as Normal, Abnormal, or Null (non-diagnostic). The platform provides a secure portal where authenticated users upload images and receive instant predictions with confidence scores.
Problem
Early oral cancer detection depends on timely visual assessment, but manual screening is slow and not always accessible. The project aimed to build a lightweight AI-assisted screening tool that helps flag potentially abnormal cases for further clinical review.
Features
Outcome
Validation accuracy
0
Macro F1-score
0
Confidence threshold
0
Growth signal
Figures animate in as you scroll — pulled from this project's published results.
- Delivered a working AI-assisted oral cancer screening web application
- Enabled fast image-based classification with confidence scoring
- Provided a foundation for clinical review workflows and further model iteration
- Demonstrated end-to-end ML pipeline: dataset → training → deployment → UI
Challenges & solutions
- Class imbalance in the dataset → Careful sampling and epoch-based training monitoring
- Non-diagnostic images → Introduced the Null class for unclear/invalid inputs
- End-to-end usability → Combined model inference with a simple authenticated web portal
