Scroll to follow the model's reasoning pipeline, from input image to final prediction.
This sequence reflects how Pulmora AI processes data in practice: image ingestion, inference, Grad-CAM visualization, evaluation, and responsible interpretation.
Explainable chest imaging demo
Open AnalyzerEducational Chest Imaging Intelligence
Pulmora AI
Pulmora AI transforms chest imaging into an interpretable AI experience. Analyze X-ray images using a deep learning model, evaluate prediction confidence, and explore Grad-CAM visualizations within a clean, anatomy-inspired interface.
See beyond. Learn deeper. Inspire care.
FastAPI-powered inference pipelineGrad-CAM visual explainabilityPerformance metrics for learning and research
Interactive Analysis
Upload a chest X-ray image and explore how the model interprets it.
Disclaimer: Pulmora AI is intended for educational purposes only and is not a medical diagnostic tool.
Select a chest X-ray image (JPG/PNG)Images are processed temporarily and not stored.
Analysis Preview
PredictionAwaiting analysis
Confidence--
OverlayPending
The heatmap highlights regions influencing the model's prediction. It is an interpretability aid and does not guarantee correctness.
Interpretability Flow
Scrolltofollow
themodel
reasoningpipeline
[
Process Overview
]Upload
Predict
Explain
Measure
Interpret
Review
Feature Map
Navigate the system as a structured pipeline of decisions.
Analyze
Run inference on uploaded images and review the model's prediction confidence.
Energy94%
Model Insights
Performance metrics for learning and research.
Accuracy91%
Precision90%
Recall93%
Confusion Matrix
Normal -> Correct186
Normal -> Misclassified14
Pneumonia -> Misclassified11
Pneumonia -> Correct189
Loss Curves
Accuracy Curves
Risks & Limitations
Pulmora AI should be used as a learning tool, not a source of medical decisions.
Model Limitations
Predictions may be incorrect, particularly on data outside the training distribution.
Dataset Bias
Imbalances in data sources, imaging conditions, or annotations can affect results.
Interpretation Risks
Heatmaps indicate model attention, not clinical proof.
Future Enhancements
Planned improvements to extend system capability.
Multi-Disease Detection
Expand beyond binary classification to support broader chest pathology exploration.
PDF Reporting
Generate exportable educational summaries that pair prediction, confidence, and explanation visuals.
Guided Interpretation
Add conversational guidance that explains the output in plain language for learners.