How dhritij27/structural-damage-recognition works
This is a structural-damage image-classification project: users train classical and transfer-learning models on labeled building images, then assess predictions and compare performance. main.py owns dataset loading, preprocessing, model training, and in-process evaluation; evaluate.py provides a separate saved-model evaluation and dataset-inspection CLI. utility.py contains visualization and prediction-analysis helpers. The README documents broader capabilities, but the sampled evaluate.py imports utils (not present in the tree), so no connection from it to utility.py is asserted.
Main entry points: main.py Training Entry · evaluate.py Saved-Model Evaluation · utility.py Analysis Visualizations
Components 16
Training
- Training Entry
main.py - Load and Split
main.py - Classical Models
main.py - MobileNet Training
main.py - InceptionV3 Training
main.py - Model Metrics
main.py - Training History
main.py - Model Comparison
main.py
Evaluation
- Evaluation CLI
evaluate.py - Saved-Model Evaluation
evaluate.py - Dataset Inspection
evaluate.py
Analysis
- Analysis Visualizations
utility.py - Prediction Export
utility.py
Other components
- User
- Labeled Images
- ImageNet Weights
Connections 17
- User → Training Entry: runs training
README.md - Labeled Images → Load and Split: provides images
main.py - Training Entry → Load and Split: loads and splits
main.py - Training Entry → Classical Models: trains models
main.py - Training Entry → MobileNet Training: trains model
main.py - Training Entry → InceptionV3 Training: trains model
main.py - MobileNet Training → ImageNet Weights: loads weights
- InceptionV3 Training → ImageNet Weights: loads weights
- Training Entry → Model Metrics: evaluates models
main.py - Training Entry → Training History: plots history
main.py - Training Entry → Model Comparison: compares models
main.py - User → Evaluation CLI: runs evaluation
README.md - Evaluation CLI → Saved-Model Evaluation: dispatches evaluation
evaluate.py - Evaluation CLI → Dataset Inspection: dispatches samples
evaluate.py - Saved-Model Evaluation → Training Entry: loads classifier
evaluate.py - Saved-Model Evaluation → Analysis Visualizations
- Saved-Model Evaluation → Prediction Export