IMVIP 2026 Conference Programme
Irish Machine Vision and Image Processing Conference • Maynooth University • August 26–28, 2026 (Conference venue TSI039 Maynooth North Campus)
Day 1: Wednesday, August 26, 2026
View Abstract & Bio
Abstract: The field of computer graphics is changing fast in the era of generative AI. New generative modeling techniques are transforming multiple stages of the graphics pipeline, from image and video synthesis to 3D content generation, stylization, and material editing. This keynote will present our recent work in generative graphics, focusing on 3D generation and appearance control. We will then examine the evolving role of generative AI in computer graphics and explore the prospect of physically based generative rendering through light transport simulation in autoencoder latent spaces.
Bio: Binh-Son Hua is an Assistant Professor of Computer Science at Trinity College Dublin. His research interests are at the intersection of computer graphics, computer vision, and machine learning with a focus on generative AI in the 3D space, particularly on 3D content generation and rendering using generative AI techniques. He received his PhD in Computer Science from National University of Singapore, and spent his postdoctoral research at The University of Tokyo, Japan and Singapore University of Technology and Design. He was recognized with the Best Paper Honorable Mention award at International Conference on 3D Vision 2016, and as an Outstanding Reviewer at CVPR 2023. He has been actively serving as an area chair for CVPR, NeurIPS, BMVC, WACV, an associate editor for TVCG, a technical program committee member for SIGGRAPH Asia, and a senior program committee member for AAAI. His research is funded by Research Ireland under the Frontiers for the Future Programme.
Day 2: Thursday, August 27, 2026
- MicroBird: Embedded Bird-Audio Classification with Microjoule-Range CNN Inference on the MAX78002
- Towards Long-Term Pig Re-Identification: Investigating Human ReID Methods in Pigs
- Model Compression for Very Short-Term Solar Irradiance Forecasting Using Infrared Sky Images
- Comparative Study of Parameter-Efficient Fine-Tuning on Low-Resource Vision-Language Modelling: Burmese-CLIP
- Perspective-Aware Billboard Corner Detection with a Web-Based End-to-End System
- Automask: Promptable Concept Segmentation with SAM3 for Glacier 3D Reconstruction from UAV Imagery
- Low-Data RGB-to-NIR Translation with Augmentation and Pseudo-Label Learning
- Geo-referenced small object detection in oblique aerial imagery
- Similarity-Augmented Roadside Object Triangulation from street level images
View Abstract & Bio
Abstract: Autonomous driving has evolved from a collection of limited highway-assistance features, such as Lane Keeping Assist and Adaptive Cruise Control, to increasingly capable systems supporting urban driving, automated parking and eyes-off driving. A key driver of this transformation has been the evolution from traditional rule-based software built from millions of lines of explicitly engineered code to AI-driven architectures powered by billions of learned neural-network parameters. This shift is fundamentally changing how autonomous systems perceive, reason, plan, and act in the physical world. In this talk, Ganesh will provide a bird’s-eye view of this evolution, tracing the journey from modular autonomy architectures and end-to-end systems to emerging Vision-Language-Action Models and Physical AI systems. The talk will explore how advances in perception, sensing, representation learning, and multimodal foundation models are coming together to create increasingly general and efficient autonomy stacks.
Bio: Ganesh Sistu is leading multiple global research teams focusing on automated driving and parking at Valeo Ireland. With over 15 years of experience in computer vision and machine learning, he has contributed to 40+ publications in top-tier conferences such as ICCV, ICRA, and WACV, garnering over 2,000 citations and 30+ US and EU patent filings. He also plays a pivotal role in guiding the future of AI education as an Adjunct Assistant Professor at the University of Limerick and as an industry board member for Research Ireland's Data Science PhD and the National MSc in AI programs at the University of Limerick, blending academic insight with industry expertise.
- Efficient automatic traffic light alteration for sparse traffic flow
- Automatic Passing Vehicle to Cyclist Distance Estimation
- A Computer Vision Tool to Aid Beginner Electric Bass Players with Fretboard Navigation
- A Hybrid Deployment Framework for Industrial Quality Control Using GelSight Tactile Sensing
- Parameter-Efficient Dense Prediction for Cross-View Geo-Localisation in Dense Urban Environments
- Transformers Cross-Attention Audio-Visual Fusion
- WeldNet: Vision-Based Weld Inspection for Intelligent Manufacturing Systems
- Enhancing an Emotion-Labelled Video Dataset with Pose Estimation Key-points for Face, Body and Hands
- Bundled Feature Encoding: A Shift-Tolerant Image Representation for Hyperdimensional Computing
- KeyJoint-STGCN: Keypoint-Based Spatio-Temporal Graph Learning for Monocular Trajectory Prediction
- Multi-session Online Open-Vocabulary Object Mapping
- Towards ISP-Free Fisheye Camera Perception: An Evaluation in Image and BEV Space
- A Domain and Sensor-Adaptive Metric Road Width Estimation from Dual Rear Cameras via Transfer Learning
Day 3: Friday, August 28, 2026
- SCULPT: Training Edge Vision Models for Post-Training Quantization Readiness
- Pix2Repair: Implicit Shape Restoration from Images
- ARAS-FCWNet: Low-cost and Light-weight Model for Forward Collision Warning (FCW) System in Unstructured ARAS
- Vision-Based Robotic Hot-Water Weeding for Precision Agriculture
- Training-Free Geometric Stabilization for Monocular 3D Detection in Advanced Rider-Assistance Systems
- Diffusion-Based Saliency Map Generation from RGB Images
- Thumbnail Guided Tissue Compaction (TGTC) - Whole Slide Image Preprocessing for Lossless Image Compression
- Body Mass Index (BMI) estimation from a single frontal face image using a Bayesian-ViT architecture
- Are Action Units Priors or Pathways? A Diagnostic Study of Temporal Valence-Arousal Estimation
- Parameter-Efficient Adaptation of V-JEPA 2 for Egocentric Action Anticipation