Train → Explain → Improve → Prove
BNNR automatically improves your PyTorch vision models using XAI — find what your model gets wrong,
fix it with intelligent augmentation, and prove the result with structured reports and a live dashboard.
Already have a trained model? Run bnnr analyze for metrics, XAI, failure patterns, and recommendations — no retraining.
Model analysis docs
Watch the full demo (with audio) on bnnr.dev
| Step | What happens |
|---|---|
| Train | Start with your PyTorch model and data. BNNR trains a baseline, then iteratively evaluates candidate augmentations — keeping only those that measurably improve performance. |
| Explain | OptiCAM, GradCAM, NMF, and CRAFT saliency maps reveal what the model focuses on. Per-class diagnoses expose blind spots and biases invisible to accuracy alone. |
| Improve | Intelligent Coarse Dropout (ICD) masks salient regions, forcing the model to learn from context. AICD sharpens focus on key features. Both are XAI-driven and automatic. |
| Prove | A structured report with metrics, XAI heatmaps, branch decisions, and before/after comparisons — shareable, auditable, and ready for stakeholders. |
- Zero-config CLI —
bnnr trainandbnnr quickstartwork without a YAML file; sensible defaults built in - Model analysis (
bnnr analyze) — full diagnostic report on any trained checkpoint without retraining - Auto-Augment Search — iterative branching strategy that tests augmentations against a baseline
- Image Classification & Object Detection — classification, multi-label, and detection (COCO-mini / YOLO) with bbox-aware augmentations and mAP metrics (v0.3.1)
- XAI Explainability — OptiCAM, GradCAM, NMF, and CRAFT heatmaps with per-class severity and trend analysis
- ICD & AICD — XAI-driven augmentations that use saliency maps to mask or focus image regions
- Real-Time Dashboard — live monitoring with branch trees, metrics, XAI previews; mobile via QR code
- Auditable Reports — structured JSON + static dashboard export for stakeholders
pip install "bnnr[dashboard]"
python3 -m bnnr train --dataset cifar10 --preset light --with-dashboardInteractive wizard:
python3 -m bnnr quickstartAnalyze an existing checkpoint:
python3 -m bnnr analyze --model checkpoints/best.pt --data cifar10 --output ./analysis_outPython API (advanced):
from bnnr import quick_run, BNNRConfig
result = quick_run(model, train_loader, val_loader, config=BNNRConfig(m_epochs=5, max_iterations=3, device="auto"))
print(result.best_metrics)| Repository | Description |
|---|---|
bnnr |
Core library — CLI, training, augmentations, XAI, dashboard |
bnnr-website |
Site at bnnr.dev — docs + demo video with audio |
| Demo | Try it |
|---|---|
| Classification (STL-10) |  |
| Multi-Label Classification | |
| Augmentations Guide | |
| Object Detection (YOLO) | |
| Bring Your Own Data | |
Mateusz Walo — Founder & Lead Developer Architect behind BNNR's core engine, XAI pipeline, and model improvement loop.
Diana Morzhak — Software Developer & QA Engineer Feature development, quality assurance, and end-to-end testing.
Dominika Zydorczyk — Community & Communications Specialist Community outreach, content strategy, and project awareness.
Zuzanna Saczuk — Graphic Designer & Brand Lead Visual identity — logo, neon branding, UI, and assets.
bnnr.dev • PyPI • GitHub • Discussions • Issues
MIT License © 2025–2026 BNNR Team