Video/Audio Quality in Computer Vision Workshop @ WACV2022

Many machine learning tasks and computer vision algorithms are susceptible to video/audio quality artifacts. Nonetheless, most visual learning and vision systems assume high-quality video/audio as input. In reality, noises and distortions are common in video/audio capturing and acquisition process. Oftentimes, artifacts can be introduced in the video compression, transcoding, transmission, decoding, and/or rendering process. All of these quality issues play a critical role on the performance of learning algorithms, systems and applications, therefore could directly impact the customer experience.

Many practical applications face the challenge of understanding video/audio quality. A few specific examples are:

• Streaming services such as Netflix, Prime Video, Hulu, and so on, need to inspect and assess the video and audio quality throughout the process to ensure high Quality of Experience (QoE);

• At the core of autonomous driving is advanced computer vision algorithms to detect objects (such as pedestrians, cars, traffic lights, etc.) on the road in real time. The robustness of such algorithms under edge cases(such as in the presence of image distortions, under low-light conditions, or severe weather conditions) is key to the wide spread application of the technology;

• Cashier-less grocery stores, with Amazon Go being a pioneer in the industry, need high quality video to build accurate tracking algorithms, to ensure a fast, reliable, and seamless customer experience.

As such, video/audio quality has become increasingly important in computer vision products, systems, and services, yet has not received enough attention in the general machine learning and computer vision community. Therefore, it is critical to systematically investigate learning performance on input videos and/or audio with variegated quality issues (noises, distortions, and artifacts, etc.). Interesting investigative topics include: determining the source of video/audio distortions and artifacts, how compression and transcoding processes affect the video/audio characteristics, how to assess video/audio quality effectively, how video/audio quality issues affect various visual learning tasks, and what we can do to address quality issues to suit our visual learning use cases.

In addition, it is equally important to leverage the advances in learning to improve the state-of-the-art video/audio quality assessment technologies, as well as developing new learning-based video/audio quality improvement algorithms and applications. We especially welcome new ideas and contributions that embrace the video/audio quality in visual learning and drive new video/audio assessment and mitigation techniques via latest development in machine learning, deep learning, and computer vision.

This workshop addresses topics related to video/audio quality in machine learning and computer vision. The topics include, but are not limited to:

• Evaluation of video/audio quality in machine learning and computer vision use-cases such as object detection, segmentation, tracking, and recognition;

• Analyze, model and learn the quality impact from video/audio acquisition, compression, transcoding, transmission, decoding, rendering, and/or display;

• Novel video/audio quality assessment methodologies: full reference, reduced-reference, and non-reference;

• Video/audio quality issues in synthesized and/or computer-generated video/audio data;

• Techniques used to remove artifacts such as shadows, glare, and reflections, etc.;

• Techniques used to improve quality such as brightening, color adjustment, sharpening, inpainting, deblurring, denoising, de-hazing, de-raining, demosaicing, etc.;

• Video/image quality improvement on resolution, frame rate, color gamut, dynamic range (SDR vs. HDR), blurring, noise, lighting, etc.;

• Datasets, statistics, and theory of video/audio quality;

• Research, applications and system development of the above.