| UTC+11 | UTC+9 | UTC+7 | UTC+5:30 | UTC+5 | UTC+3 | UTC+2 | UTC+1 | UTC-3 | UTC-4 | UTC-5 | UTC-6 | UTC-7 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2300 | 2100 | 1900 | 1730 | 1700 | 1500 | 1400 | 1300 | 0900 | 0800 | 0700 | 0600 | 0500 |
| 2400 | 2200 | 2000 | 1830 | 1800 | 1600 | 1500 | 1400 | 1000 | 0900 | 0800 | 0700 | 0600 |
| +0100 | 2300 | 2100 | 1930 | 1900 | 1700 | 1600 | 1500 | 1100 | 1000 | 0900 | 0800 | 0700 |
| +0200 | 2400 | 2200 | 2030 | 2000 | 1800 | 1700 | 1600 | 1200 | 1100 | 1000 | 0900 | 0800 |
Notes
1. Time in Halifax, Nova Scotia, Canada is UTC-3.
2. Time presented does not take into account daylight saving time.
| UTC-3 | UTC+7 | Oct 10 | Oct 11 | Oct 12 | ||||
|---|---|---|---|---|---|---|---|---|
| 9am | 7pm | Keynote | Halifax | Keynote | ||||
| 10am | 8pm | T1S1 | T2S1 | T3S1 | T1S1 | T2S1 | T3S1 | |
| 11am | 9pm | |||||||
| 12pm | 10pm | T1S2 | T2S2 | T3S2 | T1S2 | T2S2 | T3S2 | |
| 1pm | 11pm | |||||||
On-line session
On-site session
On-line sessions (October 10 and 12) all run through a single Zoom meeting. Join with the link below, then move into the breakout room matching your track for parallel sessions; the main room is used for the Keynote.
| Zoom Link | Join Zoom Meeting |
|---|---|
| Meeting ID | 817 7330 3991 |
| Passcode | 490975 |
| Track | Breakout Room |
|---|---|
| Track 1 | Breakout Room 1 |
| Track 2 | Breakout Room 2 |
| Track 3 | Breakout Room 3 |
| Track | Session Chair | Halifax Time (UTC-3)--:--:-- | Author local time View Start/End in: | ID | Article Title | |
|---|---|---|---|---|---|---|
| Start | End | |||||
| Track 1Trustworthy Medical Computer Vision and Clinical Retrieval-Augmented Generation (RAG) | Sudheer Reddy D ADRIN, ISRO India |
1000 | 1020 | UTC+7 | 151 | A Monte Carlo Dropout Framework for Uncertainty-Aware Endoscopic Depth Estimation |
| 1020 | 1040 | UTC+7 | 122 | A Two-Branch Hybrid Model for Imbalanced Lumbar Disc Grading and Zero-Shot Label Extension | ||
| 1040 | 1100 | UTC+7 | 191 | Generative AI–Driven Brain Tumor Segmentation with Counterfactual Explainability via Healthy Brain Reconstruction | ||
| 1100 | 1120 | UTC+7 | 155 | KAN-SegFormer: Kolmogorov-Arnold Decoding for Brain Tumor Segmentation | ||
| 1120 | 1140 | UTC+5:30 | 66 | Kernel Density Guided Semantic Context Boundary Smoothing for Hallucination-Aware Retrieval Augmented Generation | ||
| 1140 | 1200 | UTC+2 | 157 | LiteMedRAG: Selective Retrieval-Augmented Compact Medical Visual Question Answering | ||
| Track 2Autonomous Systems, Trajectory Prediction & Planning | Mariusz Wzorek Linköping University Sweden |
1000 | 1020 | UTC+11 | 77 | C-TAG: Context-aware Temporal Attention Graph for Trajectory Prediction |
| 1020 | 1040 | UTC+7 | 81 | A Neural Heuristic-Enhanced Incremental Planner for Mobile Robot Navigation in Dynamic Environments | ||
| 1040 | 1100 | UTC+7 | 173 | Hybrid Semantic Positioning via Sim-to-Real Transfer: Leveraging Digital Twins for Constrained Edge Localization | ||
| 1100 | 1120 | UTC+7 | 192 | VLA-Traj: An Analytically Bounded FPGA Accelerator for Deterministic Transformer–GRU Trajectory Planning | ||
| 1120 | 1140 | UTC+7 | 144 | AURORA: Robust Interactive Left-Turn Planning at Unsignalized Intersections for Autonomous Vehicles with Uncertainty and Non-Holonomic Curvature Awareness | ||
| 1140 | 1200 | UTC+2 | 177 | NEREUS: A Modular Probabilistic Framework for Context-Aware Ship Trajectory Forecasting | ||
| Track 3Edge Intelligence & Lightweight Computer Vision | Ravi Shankar G CVR College of Engineering India |
1000 | 1020 | UTC+7 | 115 | LIPE: Lightweight Landmark-Guided Image Patch Embedder for Real-Time Edge Gaze Estimation |
| 1020 | 1040 | UTC+7 | 136 | Placement Matters: An Accuracy-Cost Study of Attention Modules for Lightweight UAV Person Detection | ||
| 1040 | 1100 | UTC+7 | 195 | Multi-Level Knowledge Distillation for Lightweight Image Captioning | ||
| 1100 | 1120 | UTC+7 | 67 | Detector Selection as a Network Science Problem: Accuracy–Scalability Trade-offs in Distributed Edge Surveillance | ||
| 1120 | 1140 | UTC+2 | 183 | AI-Enabled IoT Smart Shelf System Design for SME Real-Time Inventory Monitoring and Replenishment Decision Support | ||
| 1140 | 1200 | UTC+2 | 99 | CMF-Lite: Lightweight Cross-Cue Fusion for Generalizable Deepfake Video Detection | ||
| Track | Session Chair | Halifax Time (UTC-3)--:--:-- | Author local time View Start/End in: | ID | Article Title | |
|---|---|---|---|---|---|---|
| Start | End | |||||
| Track 1Agricultural Vision & Disease Detection | S Suguna Mallika CVR College of Engineering India |
1200 | 1220 | UTC+7 | 27 | Feature Refinement via C2f-Attention: An Enhanced YOLOv8 Approach for Tomato Ripeness Detection |
| 1220 | 1240 | UTC+7 | 57 | SACTTNet: A Hybrid Convolutional-Transformer Architecture with Spatially-Aware Convolutional Tokenization for Improved Bean Disease Classification | ||
| 1240 | 1300 | UTC+7 | 82 | Sobel-Enhanced RGB Feature Integration for Plant Leaf Disease Classification | ||
| 1300 | 1320 | UTC+7 | 142 | MORPH-YOLO: A High-Frequency Morphology-Preserving Neck for Fine-Grained Pest Detection | ||
| 1320 | 1340 | UTC+7 | 147 | Dynamic Semantic Prototype Adapter for Few-Shot Plant Disease Recognition | ||
| 1340 | 1400 | UTC+2 | 139 | Few-Shot Learning for Low-Resource Natural Language Inference for French | ||
| Track 2Robust Visual Restoration & Context-Aware Analytics | Yoke Leng Yong Hong Kong College of Technology (HKCT) Hong Kong |
1200 | 1220 | UTC+7 | 44 | SPFormer: A Physics-Guided Spectral Modeling with Adaptive Spectral Decomposition for Underwater Image Enhancement |
| 1220 | 1240 | UTC+7 | 54 | Towards Robust Scale Consistency in Self-Supervised Monocular Video Depth Estimation | ||
| 1240 | 1300 | UTC+7 | 86 | HybridLLIENet: Multi-Space Luminance Prior Fusion for Low-Light Image Enhancement | ||
| 1300 | 1320 | UTC+7 | 194 | CPGA-Stable: Improving Optimization Stability in Physics-Inspired Low-Light Image Enhancement | ||
| 1320 | 1340 | UTC+7 | 107 | When Resolution Beats Slicing: Training-Free Inference Policies for Aerial Person Detection | ||
| 1340 | 1400 | UTC-5 | 52 | Comparative Analysis of Native and Translation-Based Multimodal VQA Systems for the Bangla Language | ||
| Track 3Trustworthy Small Language Models and Lightweight Deep Learning | Haji Gul City University of Science and Information Technology Pakistan |
1200 | 1220 | UTC+7 | 198 | Resource-Efficient Knowledge Graph Question Answering with Small Language Models |
| 1220 | 1240 | UTC+7 | 141 | NFTransUNet: A Neuro-Fuzzy Deep Learning Model for Breast Tumor Segmentation in Big Data Environment | ||
| 1240 | 1300 | UTC-7 | 119 | Knowing When to Defer: A Risk-Calibrated Deferral Policy for Agentic AI in Financial Decision-Making | ||
| 1300 | 1320 | UTC+7 | 78 | CLM-Net: A Channel-Optimized Hybrid Network for Varroa Mite Segmentation | ||
| 1320 | 1340 | UTC+2 | 133 | InjectGuard-RAG: Lightweight Defence Against Indirect Prompt Injection in Retrieval-Augmented Small-LLM Agents | ||
| 1340 | 1400 | UTC-6 | 196 | Provable Limits and Certified Deferral for Verbalized Uncertainty in Small Language Models | ||
No parallel online paper tracks are scheduled on this day.
| Item | Details |
|---|---|
| Asynchronous Paper Hosting | All conference papers hosted asynchronously online, available from Saturday, October 10, 2026 onwards. |
| Asynchronous Industry Showcase | Videos, demos, and posters, available from Saturday, October 10, 2026 onwards. |
| Welcome Reception | Saturday, October 10, 2026 | 12:00 PM–2:00 PM (Halifax Time, ADT)
Venue: TBA |
| In-Person Technical Session | Sunday, October 11, 2026 | 9:00 AM–11:00 AM (Halifax Time, ADT)
MCDA Learning Suite, Science Building Room S126, Saint Mary's University, Halifax, NS B3H 3C3 Chair: Andrew Fisher, University of New Brunswick, Fredericton |
| Conference Banquet | Sunday, October 11, 2026 | 12:00 PM–2:00 PM (Halifax Time, ADT)
Venue: TBA |
Sunday, October 11, 2026 | 9:00 AM–11:00 AM (Halifax Time, ADT)
MCDA Learning Suite, Science Building Room S126, Saint Mary's University, Halifax, NS
Session Chair: Andrew Fisher, University of New Brunswick, Fredericton
Six selected papers from MIWAI 2026, covering anomaly detection, speech classification, possibilistic clustering, computational humour, sensor fusion, and representation learning. Each presentation is allocated 20 minutes total (15 minutes presentation + 5 minutes Q&A).
| Session | Session Chair View Start/End in: | Halifax Time (UTC-3)--:--:-- | Presenter local time | ID | Title | |
|---|---|---|---|---|---|---|
| Start | End | |||||
| In-Person SessionMCDA Learning Suite, Science Building Room S126, Saint Mary's University, Halifax, NS | Andrew Fisher University of New Brunswick, Fredericton |
0900 | 0920 | UTC-3 | 49 | A Lightweight Compression-Driven Method for Multi-Scale Image Anomaly Detection |
| 0920 | 0940 | UTC-3 | 95 | Injecting Label Structure into Embedding Space for Ordinal and Circular Speech Classification | ||
| 0940 | 1000 | UTC-3 | 165 | A Possibilistic K-Modes Framework for Decremental Object Processing | ||
| 1000 | 1020 | UTC-3 | 166 | On assessing and repairing jokes in a logical framework | ||
| 1020 | 1040 | UTC-3 | 167 | Attention-Based Fusion of Homogeneous Sensors for Robust Classification Across Varying SNRs | ||
| 1040 | 1100 | UTC-3 | 190 | From Local Density to Lexical Features: A Cross-Scale Study of Unembedding Spaces | ||
Notes for presenters
Full Halifax program documents:
12 partners · 10 sectors · Asynchronous, available from Saturday, October 10, 2026 onwards
The MIWAI 2026 Industry Showcase brings together companies and public-sector partners applying artificial intelligence to real-world problems across Nova Scotia and beyond. Each partner presents their work in three parts — a recorded presentation, an infographic summarising the project at a glance, and a written abstract — open to explore at your own pace, before, during and after the conference. Explore the showcase at ml.cs-smu.ca/miisc.
| # | Partner & Sector | Project |
|---|---|---|
| 1 | Artinus Consulting Inc. Supply Chain & Logistics |
SKU Pilot: An ERP-Connected AI Control Tower Forecasting, 3D container loading and landed-cost tools that help import-intensive businesses move from spreadsheets to proactive planning. |
| 2 | Halifax Learning Education |
Automating Phonological Assessments A scalable platform that turns one-to-one phoneme testing into an automated, asynchronous assessment with educator verification. |
| 3 | Nova Scotia Health Public Sector |
Strategic Resource Optimization & Decision Support
Personalised dashboards that align regional facility resources with community demand — and train 15 Master's interns in digital health. |
| 4 | Kardio Diagnostix Healthcare AI |
AI-Powered Pediatric Heart Murmur Triage Physics-informed signal processing and deep learning to distinguish benign murmurs from pathologic abnormalities and cut unnecessary referrals. |
| 5 | Wellnify.ai Health & Wellness |
Holistic Personalization for Wellness Engagement Engagement dashboards plus an NLP recommendation engine that suggests tailored Mind, Body and Spirit activities for healthier communities. |
| 6 | MacDon Industries Ltd. Agriculture |
Synthetic Data for Agricultural Segmentation Photorealistic synthetic data and domain adaptation to train crop-row semantic segmentation models with far less manual labelling. |
| 7 | Hatfield Consultants LLP Environmental Monitoring |
Automated Linear Disturbance Mapping Unsupervised domain adaptation maps roads and corridors in boreal forest from Sentinel-2 satellite imagery, supporting caribou conservation. |
| 8 | R-Tech Industries Agriculture |
The R-Tech Data Rover: Prototype to Product A self-propelled, GPS-guided rover collecting plant images and environmental data in test plots, organic farms and commercial fields. |
| 9 | TACKLE Lab Construction |
AI-Driven Pre-Planning for Construction From 2D drawings to 3D models (BIRD), material forecasts (JACK) and optimised panel layouts (PAAD) that minimise manufacturing waste. |
| 10 | WorkCore AI Software & Platforms |
An Enterprise AI Agent Workforce Secure, governed AI agents using retrieval-augmented generation to automate knowledge-intensive work for small and medium-sized businesses. |
| 11 | Invisipay Inc. Retail |
Building Autonomous, Checkout-Free Retail Camera- and AI-powered stores that track interactions in real time, update carts automatically and process secure payment on exit. |
| 12 | Delmore “Buddy” Daye Learning Institute Education |
Evidence-Driven Africentric Strategies Africentric Cultural Intelligence with multilevel modelling and machine learning to improve schooling outcomes for learners of African descent. |
Showcase Committee
Showcase documents:
| Track | Session Chair | Halifax Time (UTC-3)--:--:-- | Author local time View Start/End in: | ID | Article Title | |
|---|---|---|---|---|---|---|
| Start | End | |||||
| Track 1Multimodal & Medical AI Applications | Sathya Babu Korra IIIT-DM, Kurnool India |
1000 | 1020 | UTC+7 | 70 | SlimFusion: Lightweight Audio–Visual Emotion Recognition for Edge Inference |
| 1020 | 1040 | UTC+7 | 143 | ICU Readmission Prediction for Patients with Alcohol-related Disorders Based on Multimodal Deep Learning | ||
| 1040 | 1100 | UTC+7 | 148 | Overcoming Boundary Uncertainty in Semi-Supervised Cardiac Segmentation via Masked Image Modeling and Multi-Scale Consistency | ||
| 1100 | 1120 | UTC+5:30 | 97 | Learning Joint Representations through Adaptive Late Fusion for Enhanced Tuberculosis Detection from Chest Radiographs | ||
| 1120 | 1140 | UTC+2 | 132 | Multi-Scale Visual Feature Extraction for Medical Visual Question Answering | ||
| 1140 | 1200 | UTC+2 | 110 | Calibration-Aware Lightweight Model Selection for Trustworthy Skin Lesion Classification | ||
| Track 2Knowledge Graphs and Information Retrieval | Naveen Nekuri University of Hyderabad India |
1000 | 1020 | UTC+7 | 160 | Multi View LightGCN over a Human Resource Knowledge Graph for Personalised Internal Career Development Recommendation |
| 1020 | 1040 | UTC+7 | 152 | LOKG-IR: An Open and Lightweight Knowledge Graph Framework for Semantic Image Retrieval | ||
| 1040 | 1100 | UTC+7 | 169 | Attentional Regression Approach Incorporating User-based and Review-based Explanations for Recommendation Systems | ||
| 1100 | 1120 | UTC+7 | 138 | DEQ-EC: Dynamic Exposure-Aware Measurement of Echo Chambers in Social Media Information Networks | ||
| 1120 | 1140 | UTC+1 | 73 | Drug Repositioning through Interpretable Link Prediction in Drug-Disease Networks | ||
| Track 3Trustworthy AI, Cybersecurity & IoT Intelligence | Dayang Norhayati Abang Jawawi Universiti Teknologi Malaysia Malaysia |
1000 | 1020 | UTC+7 | 121 | A Physics-Informed and Energy-Based Framework for Countering Micro-Doppler Spoofing in UAV Radar Signatures |
| 1020 | 1040 | UTC+7 | 137 | Rank-Collapse and Detail Re-injection for Efficient Safety-Helmet Detection | ||
| 1040 | 1100 | UTC+5 | 197 | When Identical Conditions Yield Different Authority: Measuring Path-Dependence in Trust-to-Authorization Policies | ||
| 1100 | 1120 | UTC+2 | 69 | Hybrid Feature Selection with a Meta-Learner for Explainable IoT Intrusion Detection | ||
| 1120 | 1140 | UTC-4 | 51 | Stability-Aware Model Selection for Explainable Artificial Intelligence: Accuracy, Consistency, and Generalization | ||
| 1140 | 1200 | UTC-7 | 92 | Data-Driven Intrusion Detection for UAV Networks: A Machine-Learning System for Multi-Attack Recognition | ||
| Track | Session Chair | Halifax Time (UTC-3)--:--:-- | Author local time View Start/End in: | ID | Article Title | |
|---|---|---|---|---|---|---|
| Start | End | |||||
| Track 1LLMs, NLP & Intelligent Document Processing | Sindhura Gade Principal Software Engineer, OneTrust USA |
1200 | 1220 | UTC+7 | 100 | LLM-Based Code Generation for Android Automotive OS Vehicle Hardware Abstraction Layer |
| 1220 | 1240 | UTC+7 | 124 | An Adversarial Dual-Component Framework for Toxic Speech Detection with LLM-GAN Augmentation and Toxicity-Conditioned Cross-Layer Attention | ||
| 1240 | 1300 | UTC+7 | 9 | Enhancing YOLOv8 with Symmetric DINO for Low-Data Brain Tumor Detection | ||
| 1300 | 1320 | UTC+5:30 | 68 | Sentiment-Aware Semantic Graph Construction for Harmful Tweet Classification using Graph Neural Networks | ||
| 1320 | 1340 | UTC+2 | 48 | CART-Doc: A Compact Design Space for Confidence-Routed Document Key Information Extraction | ||
| 1340 | 1400 | UTC+2 | 56 | Pareto Optimal Counterfactual Explanations with an LLM Validation | ||
| Track 2Optimization, Reasoning & Formal AI | Tho Quan HCMUT Vietnam |
1200 | 1220 | UTC+3 | 149 | Reasoning about Transitive Variably Polyadic Relations |
| 1220 | 1240 | UTC+3 | 111 | Ranking Sparse Candidate Sets: Phase Transitions in Learned TSP Graph Pruning | ||
| 1240 | 1300 | UTC+2 | 117 | Symmetry Mappings of Causality Structures for Causal AI Formalised in a Galois Extension of Commutative Fields and Reified in a Noncommutative Field Extension | ||
| 1300 | 1320 | UTC-5 | 31 | Solving the Faculty Timetabling Problem via the Quantum Approximate Optimization Algorithm | ||
| 1320 | 1340 | UTC-6 | 60 | The Geometry of Consequence: Modeling Irreversible Multi-Agent Systems via Co-Structures | ||
| Track 3Efficient Learning, Optimization & Intelligent Systems | Hoang Anh Pham Ho Chi Minh City University of Technology (HCMUT), VNU-HCM Vietnam |
1200 | 1220 | UTC+9 | 172 | Proposal of a Rule Table Method for Classification Incorporating Naive Bayes and a Comparative Study with Existing Methods |
| 1220 | 1240 | UTC+7 | 30 | A Framework for Time-Series Anomaly Detection with Mamba Backbone | ||
| 1240 | 1300 | UTC+7 | 80 | PseuAgg: Enhancing Efficiency and Accuracy in Decentralized Federated Learning with Pseudo Aggregation | ||
| 1300 | 1320 | UTC+7 | 182 | TKEO-Guided Adaptive Pre-Emphasis for Acoustic Fish Feeding Intensity Classification | ||
| 1320 | 1340 | UTC+5:30 | 50 | Progressive Multi-Stage Evolutionary NAS Optimisation of Vision Transformers for Landslide Hazard Prediction: A PSO–NAS–Bayesian Pipeline for Multi-Spectral Remote Sensing | ||