In the past decades, digital cameras have witnessed the emergence of smart imaging solutions driven by AI and CV algorithms. The recent development of deep learning has radically improved the capacity of computational imaging. Unfortunately, how to explore the benefit of high-performance imaging, so as to inspire and promote more capable computer vision solutions has not been investigated extensively. The gap between computational imaging and computer vision remains unsolved. For example, while the latest progress of array cameras can realize gigapixel videography capability for large-scale dynamic scenes, existing computer vision tasks remain working on the traditional low-resolution visual data.
Given the popularity of high-performance imaging enabled by the array cameras even in our mobile phones, what are the opportunities and challenges for Computer Vision research? How can computer vision algorithms further help with the array camera to achieve REAL smart? They are open questions demanding prompt solutions. Given this, the 3rd GigaVision Workshop aims to bring the attention from both the computer vision and computational imaging communities.
- Semantic labeling, object detection and recognition
- Long-term multi-target visual tracking
- Action and activity recognition
- Gigapixel Imaging
- Gigapixel Videography
- Efficient network architecture design
- Prediction, modeling and analysis of large-scale scenes
- Large-scale 3D reconstruction using multiple gigapixel images
- Diffractive Deep Neural Network for Saliency Detection
- Deep optics: end-to-end optimization for intelligent camera