The advantages of collecting images from outdoor camera platforms, like UAVs, surveillance cameras and outdoor robots, are evident and clear. For instance, man-portable UAV systems can be launched from safe positions to survey difficult or dangerous terrain, acquiring hours of video without putting human lives at risk. What is unclear is how to automate the interpretation of these images - a necessary measure in the face of millions of frames containing artifacts unique to the operation of the sensor and optics platform in outdoor, unconstrained, and usually visually degraded environments.
Continuing the success of the 1st UG2 Prize Challenge workshop held at CVPR 2018, UG2+ provides an integrated forum for researchers to review the recent progress of handling various adverse visual conditions in real-world scenes, in robust, effective and task-oriented ways. Beyond the human vision-driven restorations, we also extend particular attention to the degradation models and the related inverse recovery processes that may benefit successive machine vision tasks. We embrace the most advanced deep learning systems, but are still open to classical physically grounded models, as well as any well-motivated combination of the two streams. The workshop will consist of four invited talks, together with peer-reviewed regular papers (oral and poster), and talks associated with winning prize challenge contributions.
Original high-quality contributions are solicited on the following topics:
- Novel algorithms for robust object detection, segmentation or recognition on outdoor mobility platforms, such as UAVs, gliders, autonomous cars, outdoor robots, etc.
- Novel algorithms for robust object detection and/or recognition in the presence of one or more real-world adverse conditions, such as haze, rain, snow, hail, dust, underwater, low-illumination, low resolution, etc.
- The potential models and theories for explaining, quantifying, and optimizing the mutual influence between the low-level computational photography (image reconstruction, restoration, or enhancement) tasks and various high-level computer vision tasks.
- Novel physically grounded and/or explanatory models, for the underlying degradation and recovery processes, of real-world images going through complicated adverse visual conditions.
- Novel evaluation methods and metrics for image restoration and enhancement algorithms, with a particular emphasis on no-reference metrics, since for most real outdoor images with adverse visual conditions it is hard to obtain any clean “ground truth” to compare with.
Awarded in prizes
Can the application of image enhancement and restoration algorithms as a pre-processing step improve image interpretability for automatic visual recognition to classify scene content?
The UG2+ Challenge seeks to advance the analysis of "difficult" imagery by applying image restoration and enhancement algorithms to improve analysis performance. Participants are tasked with developing novel algorithms to improve the analysis of imagery captured under problematic conditions.
Video object classification and detection from unconstrained mobility platforms
Image restoration and enhancement algorithms that remove corruptions like blur, noise and mis-focus, or manipulate images to gain resolution, change perspective and compensate for lens distortion are now commonplace in photo editing tools. Such operations are necessary to improve the quality of images for recognition purposes. But they must be compatible with the recognition process itself, and not adversely affect feature extraction or decision making.
Object Detection in Poor Visibility Environments
The performances of visual sensing and understanding algorithms can be largely jeopardized by various challenging conditions in unconstrained and dynamic degraded environments, e.g., moving platforms, bad weathers, and poor illumination. While most current vision systems are designed to perform in environments where the subjects are well observable without (significant) attenuation or alteration, a dependable vision system must reckon with the entire spectrum of complex unconstrained outdoor environments.
- Object Detection Improvement on Video
- Object Classification Improvement on Video
- (Semi-)Supervised Object Detection in Haze Conditions
- (Semi-)Supervised Face Detection in Low Light Conditions
- Zero-Shot Object Detection with Raindrop Occlusions
Al Bovik holds the Cockrell Family Endowed Regents Chair in Engineering at The University of Texas at Austin, where he is Director of the Laboratory for Image and Video Engineering (LIVE).
Lars Ericson is currently a Program Manager at the Intelligence Advanced Research Project Activity ( IARPA ), within the Office of the Director of National Intelligence, with a focus on biometrics, computer vision, sensors, and nanotechnology.
Heesung Kwon is a Senior Researcher and the Image Analytics Team lead at the U.S. Army Research Laboratory (ARL).
Manmohan Chandraker is an Assistant Professor at the CSE department of the University of California, San Diego and is the Computer Vision Lead at NEC Labs America.