Muhammad Rashid
Machine Learning Researcher specializing in Explainable AI (XAI), Computer Vision, and Visual Anomaly Detection for real-world and safety-critical systems.
Research Focus
- Explainable AI for trustworthy decision-making
- Visual anomaly detection in industrial and robotic environments
- Pixel-level feature attribution using Shapley-based methods
- Robust and efficient explanation methods for computer vision
📌 AAAI 2026: ShapBPT — Image Feature Attribution using Data-Aware Binary Partition Trees
📌 ICPE 2026 / QualITA Workshop: ShapBPT in Perspective: A Consolidated Review and an eXplainable Anomaly Detection Case Study
Systems & Engineering
- Real-time anomaly detection for robotics safety
- ROS2-based live inference pipeline
- VAE / VAE-GAN models for industrial anomaly detection
- Multi-camera monitoring and safety-area segmentation
- End-to-end workflow: training → threshold calibration → live deployment
Achieved:
- ~12 FPS real-time inference
- ~99.6% detection accuracy in industrial scenarios
- Explainable anomaly maps for safety-critical decision support
Open Source & Tools
- ShapBPT — AAAI 2026 explainability framework
- XAD — Explainable anomaly detection with ShapBPT
- LIME Stratified Sampling — Improved explanation stability
- AI on Edge Devices — Lightweight deployment pipelines
- XAI evaluation tools for saliency, attribution, and benchmarking
Selected Publications
ShapBPT: Image Feature Attribution using Data-Aware Binary Partition Trees
AAAI 2026
PDF arXiv Code Tests PyPI Docs User Study Poster
ShapBPT in Perspective: A Consolidated Review and an eXplainable Anomaly Detection Case Study
QualITA Workshop, ICPE 2026
Using Stratified Sampling to Improve LIME Image Explanations
AAAI 2024
Can I Trust My Anomaly Detection System?
XAI World 2024
Featured Projects
🔧 Real-Time Anomaly Detection for Robotics
- Industrial safety monitoring system
- Multi-area detection: RoboArm, Conveyor Belt, Pallet Left, Pallet Right
- ROS2 integration with live inference and anomaly scoring
- Visual anomaly maps for decision interpretation
ShapBPT
- AAAI 2026 explainability method for images
- Uses data-aware Binary Partition Trees
- Provides hierarchical Shapley-based feature attributions
- More image-structure-aware than standard SHAP partition explainer
PDF arXiv Code Tests PyPI Docs User Study Poster
XAD: Explainable Anomaly Detection
- Applies ShapBPT to anomaly detection systems
- Explains why an image or region is considered anomalous
- Supports interpretation of black-box anomaly detection models
LIME Stratified Sampling
- Improves stability of LIME image explanations
- Reduces variance in perturbation-based sampling
- Provides more reliable local explanations
Experience
- 🎓 PhD Researcher — University of Turin, Italy (details here)
- 🏭 Industrial Research — RuleX Innovation Labs
- 🇪🇺 EU Project — DistriMuSe: Robotics Safety & AI
- 🌐 Visiting Researcher — University of Granada, Spain (details here)
Published Python Packages
Python packages released for explainable AI and reproducible research.
| Package | Focus | Registry | Version | Links |
|---|---|---|---|---|
| lime-stratified | Stable LIME image explanations | PyPI | … | PyPI · Code |
| shap-bpt | Hierarchical Shapley image attribution | PyPI | … | PyPI · Docs |
Open Source Contributions
Selected contributions to open-source machine learning, anomaly detection, and data science software.
| Library | Organization | Focus | Stars | Contribution |
|---|---|---|---|---|
| Anomalib | Intel/Open Edge Platform | Visual Anomaly Detection | Loading... | PatchCore docs |
| FLAML | Microsoft | AutoML / ML Systems | Loading... | Anomaly detection support |
| Awesome Python for Data Science | Data-Centric AI Community | Data Science Education | Loading... | Anomaly detection tutorial |
Collaboration
I’m interested in collaborations on:
- Explainable AI and trustworthy machine learning
- Visual anomaly detection
- Industrial AI and robotics safety
- Computer vision for real-world systems
- Vision-language models and explainability
Contact
- 📧 Email:
{FIRSTNAME}.{LASTNAME}@unito.it - 🔗 GitHub: github.com/rashidrao-pk
- 📚 Google Scholar: Muhammad Rashid