RLQ 2019 : ICCV 2019 Workshop and Challenge on Real-World Recognition from Low-Quality Images and Videos
Call For Papers
What is the current state-of-the-art for recognition and detection algorithms in non-ideal visual environments? We are organizing the 2nd RLQ challenge and workshop in ICCV 2019, with an expanded scope for paper solicitation.
While the visual recognition research has made tremendous progress in recent years, most models are trained, applied, and evaluated on high-quality (HQ) visual data, such as ImageNet benchmarks. However, in many emerging applications such as robotics and autonomous driving, the performances of visual sensing and analytics are largely jeopardized by low-quality (LQ) visual data acquired from complex unconstrained environments, suffering from various types of degradations such as low resolution, noise, occlusion and motion blur. While some mild degradations may be compromised by sophisticated visual recognition models, their impacts turn much notable as the level of degradations passes some empirical threshold. Other factors, such as contrast, brightness, sharpness, and out-of-focus, all have various negative effects on visual recognition.
We organize this one-day workshop to provide an integrated forum for researchers to review the recent progress of robust recognition models from LQ visual data, and the novel image restoration algorithms. We embrace the most advanced deep learning systems, meanwhile being open to classical physically grounded models and feature engineering, as well as any well-motivated combination of the two streams.
RLQ 2019 consists of challenge, keynote speech, paper presentation, poster session, special session on privacy and ethics of person-related recognition, and a panel discussion from the invited speakers. Specifically,
This challenge is led by QMUL.
In addition to submitting the results to the online evaluation system, (Mandatory) participants are required to submit codes and fact sheet for evaluation. (Optional) They are also encouraged to submit a workshop paper.
Validation Phase: 25 June to 30 Aug
Testing Phase: 8 Aug to 30 Aug
Final Ranks Released: 30 Aug
Paper Submission Deadline: 15 Aug
Notification: 20 Aug
Camera-Ready: 28 Aug
Material Submission Deadline: 30 Aug
We will solicit full-papers from but not limited to the following topics:
*Robust recognition and detection from low-resolution image/video
*Robust recognition and detection from video with motion blur
*Robust recognition and detection from highly noisy image/video
*Robust recognition and detection from other unconstrained environment conditions
*Low-resolution image/video enhancement, especially for recognition purpose
*Image/video denoising and deblurring, especially for recognition purpose
*Restoration and enhancement of other common degradations, such as low-illumination, inclement weathers, etc., especially for recognition purpose
*Novel methods and metrics for image restoration and enhancement algorithms, especially for recognition purpose
*Surveys of algorithms and applications with LQ inputs in computer vision
*Psychological and cognitive science research with proper data processing and enhancement
*Novel calibration and registration methods on gaze, face or object images for recognition or detection purpose.
*Novel imperfect low-quality data mining, cleaning, and processing methods for training a recognition system.
*Other novel applications that robustly handle computer vision tasks with LQ inputs
*[Special Issues] Legal, privacy and ethics in person-related recognition
The accepted papers will be included in the ICCV2019 workshop CVF proceedings, and authors will be invited to present their paper either in oral or poster form.
We solicit “positioning” writeups, in the form of short non-archival abstracts. They shall address important issues that may generate a lasting impact for next 5-year research in the field of recognition in low-quality visual data. Examples may include but are not limited to,
*Proposing novel technical solutions: preliminary works and “half-baked” results are welcome
*Identifying grand challenges that are traditionally overlooked or under-explored
*Discussing rising applications where recognition from low-quality visual data might have been a critical bottleneck
*Raising new research questions that may be motivated by emerging applications
*New datasets, new benchmark efforts, and/or new evaluation strategies
*Integration of low-quality visual recognition into other research topics
The accepted abstracts will appear on the website. The workshop organizers will lead a collective positioning paper, targeted at a top-tier journal such as TPAMI or IJCV. Those whose abstracts are selected will be invited as co-authors of this paper (the author order will be alphabetical).
We will hold discussions on 'Legal, privacy, and ethics issues of person-related recognition' topic based on the previous experience from researchers, and address public concerns. This session will be led by world-renowned researchers.
We will host the panel discussion among the invited speakers and organizers. Topics TBD.