How vanithamagesh/bifam-next works
BiFAM-Next is a research pipeline for brain-tumor MRI classification and attention localization. Its main workflow prepares and partitions MRI slices, trains/evaluates the classifier (with mask-supervised attention during training), and saves predictions and metrics; analysis then examines attention and redraws paper figures. The model and local driver are sampled, while notebook workflows and several analysis capabilities are documented but not fully inspected, so those links rely on the README. Results files are retained as artifacts, not a runtime database.
Main entry points: results Predictions and metrics · run_local.py Local experiment driver · figures Paper figures · jobs_core.py Experiment jobs
Components 14
Data preparation
- Data preparation
notebooks/BiFAM_Next_jobs.ipynb
Model and training
- BiFAM-Next model
src/bifamnet.py - Experiment jobs
src/jobs_core.py - Local experiment driver
src/run_local.py - Training and evaluation
src/bifamnet.py
Evaluation and analysis
- Attention localization
notebooks/Paper1_localization.ipynb - Attention gallery
notebooks/Attention_gallery.ipynb - Ablation figure
analysis/fig_ablation.py - Ensemble figure
analysis/fig_ensemble8.py - Localization figure
analysis/fig_localization.py
Research outputs
- Predictions and metrics
results - Paper figures
figures
Other components
- Researcher
- MRI datasets
Connections 20
- Researcher → Data preparation: prepares data
README.md - MRI datasets → Data preparation: provides images
README.md - Data preparation → Training and evaluation: supplies partitions
README.md - Researcher → Experiment jobs: selects jobs
README.md - Experiment jobs → BiFAM-Next model: uses model
src/jobs_core.py - Experiment jobs → Local experiment driver: invokes driver
src/jobs_core.py - Researcher → Local experiment driver: runs experiments
README.md - Local experiment driver → BiFAM-Next model: imports module
src/run_local.py - Local experiment driver → Training and evaluation: calls training
src/run_local.py - Training and evaluation → Predictions and metrics: saves results
README.md - Predictions and metrics → Attention localization: provides checkpoints
README.md - Predictions and metrics → Attention gallery: provides checkpoints
README.md - Predictions and metrics → Ablation figure: provides metrics
analysis/fig_ablation.py - Predictions and metrics → Ensemble figure: provides predictions
analysis/fig_ensemble8.py - Predictions and metrics → Localization figure: provides localization data
analysis/fig_localization.py - Attention localization → Predictions and metrics: writes analysis
README.md - Attention gallery → Paper figures: creates figures
README.md - Ablation figure → Paper figures: writes figure
analysis/fig_ablation.py - Ensemble figure → Paper figures: writes figure
analysis/fig_ensemble8.py - Localization figure → Paper figures: writes figure
analysis/fig_localization.py