How gt-610/cv-lab-gtsrb works
This is a GTSRB traffic-sign recognition course project: users can inspect the dataset, train and evaluate classifiers, predict signs from images, compare integrated member-B models, and run a Streamlit demo. The main workflow is config-driven training over GTSRB data through model creation and training, producing checkpoints and evaluation artifacts. The graph also preserves analysis, result summarization, demo-asset preparation, and B-model artifact preparation. Edges follow shown calls, source-index references, or README statements; unsampled script internals are not assumed.
Main entry points: eval_model.py Evaluation entrypoint · train_model.py Train entrypoint · train.py Training engine · utils.py Runtime utilities
Components 19
Training and Evaluation
- Train entrypoint
scripts/train_model.py - Training configs
configs/resnet18.yaml - Evaluation entrypoint
scripts/eval_model.py
Prediction and Demo
- Single-image prediction
scripts/predict_image.py - Interactive demo
scripts/demo_streamlit.py
Dataset and Results
- Dataset analysis
scripts/analyze_dataset.py - Experiment summaries
scripts/summarize_results.py - Demo asset preparation
scripts/prepare_demo_assets.py
Member B Models
- Member B predictors
src/cvlab/b_models.py - Multi-model prediction
scripts/predict_all_models.py - B artifact preparation
scripts/prepare_b_artifacts.py
Shared ML Components
- GTSRB data loaders
src/cvlab/data.py - Class labels
src/cvlab/labels.py - Classifier models
src/cvlab/models.py - Training engine
src/cvlab/train.py - Classification metrics
src/cvlab/metrics.py - Training visualizations
src/cvlab/visualize.py - Runtime utilities
src/cvlab/utils.py
Other components
- Project user
Connections 32
- Project user → Train entrypoint: starts training
README.md - Training configs → Train entrypoint: configures
README.md - Train entrypoint → GTSRB data loaders: builds loaders
scripts/train_model.py - Train entrypoint → Classifier models: creates model
scripts/train_model.py - Train entrypoint → Training engine: runs training
scripts/train_model.py - GTSRB data loaders → Class labels: uses labels
src/cvlab/data.py - Training engine → Classification metrics: computes metrics
src/cvlab/train.py - Training engine → Training visualizations: plots results
src/cvlab/train.py - Training engine → Runtime utilities: writes run data
src/cvlab/train.py - Project user → Evaluation entrypoint: starts evaluation
README.md - Evaluation entrypoint → GTSRB data loaders: builds loaders
scripts/eval_model.py - Evaluation entrypoint → Classifier models: loads checkpoint
scripts/eval_model.py - Evaluation entrypoint → Training engine: evaluates model
scripts/eval_model.py - Evaluation entrypoint → Classification metrics: builds confusion matrix
scripts/eval_model.py - Evaluation entrypoint → Training visualizations: plots evaluation
scripts/eval_model.py - Project user → Dataset analysis: analyzes dataset
README.md - Dataset analysis → Class labels: uses labels
scripts/analyze_dataset.py - Dataset analysis → Runtime utilities: creates output directory
scripts/analyze_dataset.py - Project user → Experiment summaries: summarizes results
README.md - Experiment summaries → Runtime utilities: uses utilities
scripts/summarize_results.py - Project user → Demo asset preparation: prepares assets
README.md - Demo asset preparation → Class labels: uses labels
scripts/prepare_demo_assets.py - Demo asset preparation → Runtime utilities: uses utilities
scripts/prepare_demo_assets.py - Project user → Single-image prediction: predicts image
README.md - Single-image prediction → Classifier models: loads model
scripts/predict_image.py - Project user → Interactive demo: uses demo
README.md - Interactive demo → Classifier models: loads checkpoint model
scripts/demo_streamlit.py - Project user → B artifact preparation: prepares artifacts
README.md - Project user → Multi-model prediction: runs model comparison
README.md - Multi-model prediction → Member B predictors: runs predictions
scripts/predict_all_models.py - Member B predictors → Class labels: uses labels
src/cvlab/b_models.py - Member B predictors → Runtime utilities: selects device
src/cvlab/b_models.py