Session Topics

Advances in DNNs for perception
Approaches and advances in object tracking
Conventional (non-ML based) computer vision; detection, classification, OCR
Event/action recognition
Generative AI involving visual perception
Improving accuracy in the real world (e.g., snow, rain, fog, other noise, etc.)
Learning at the edge; few-shot learning, continuous learning
Model compression, network optimization
Monocular depth detection; ML-based, other
Perception and scene understanding using non-RGB sensors, including 3D sensors, Lidar, Radar
Practical image processing techniques; lens distortion correction, noise removal, etc
Practical neuromorphic algorithms, including spiking neural networks
Sensor fusion - vision, accoustic, heptic, etc.
Tracking
Vision transformers, including multi-modal models and use cases

A survey of practical object tracking approaches
Camera selection and image quality metrics
Ensuring quality data for deep Learning: data Collection, duration, and annotation
Fundamentals of 3D Sensors (depth sensing)
Fundamentals of optics and sensors
Fundamentals of training AI models for computer vision applications
Introduction to CNNs
Introduction to DNN-based object detectors
Introduction to ML for visual perception
Introduction to segmentation: semantic, instance, panoptic
Introduction to vision transformers
Model compression approaches – quantization, pruning, sparsity, distillation, etc.
Model training issues, fixes, insights, best practices
SLAM
Survey of object detectors and their use cases

Case studies of end-to-end computer vision based solutions, pitfalls to avoid, insights
Cloud-edge tradeoffs
End-to-end system optimization; an application is more than a neural network.
Event-based and neuromorphic sensors and processors
Image and non-image sensors and processing techniques: Lidar, Radar, ToF, stereo, audio
Industry insights and best practices in developing and deploying applications for the edge
Low-power low-cost embedded computer vision components - SoCs, CPUs, NPUs, etc...
MLOps - accuracy, precision, other performance metrics that matters in the real world and how to maintain them in production over time
MLOps - data collection, curation, annotation, generation, augmentation
MLOps - deploying and maintaining models in production
MLOps - testing a model in the lab vs in the wild, best practices and insights, explainibility
MLOps - training a model, best practices, insights
RGB sensors, optics, camera selection, image quality metrics
Standards related to computer vision based systems in general
Tools, frameworks, and techniques for developing and deploying deep neural networks
Top-to-bottom software stacks for vision-based AI systems and applications

Call for Proposals Suggested Topics