Data Challenges

Video compression standards rely heavily on eliminating spatial and temporal redundancy within and across video frames. Intra-frame encoding targets redundancy within blocks of a single video frame, whereas inter-frame coding focuses on removing redundancy between the current frame and its reference frames.

Supported by the SPS Challenge Program.

Accurate analysis of liver vasculature in three dimensions (3D) is essential for a variety of medical procedures including computer-aided diagnosis, treatment planning, or pre-operative planning of hepatic diseases.

Speech-enabled systems often experience performance degradation in real-world scenarios, primarily due to adverse acoustic conditions and interactions among multiple speakers. Enhancing the front-end speech processing technology is vital for improving the performance of the back-end systems. 

This challenge will perform the first comprehensive benchmark of the impact of a wide range of distortions on the performance of current object detection methods. The proposed database contains, in addition to the conventional real distortions, other synthesized photo-realistic distortions corresponding to real and very frequent scenarios often neglected in other databases despite their importance. The synthetic distortions are generated according to several types and severity levels with respect to the scene context.

Associated SPS Event: IEEE ICIP 2022 Grand Challenge

High Dynamic Range (HDR) imaging provides the ability to capture, manipulate and display real-world lighting. This is a significant upgrade from Standard Dynamic Range (SDR) which only handles up to 255 luminance values concurrently. While capture technologies have advanced significantly over the last few years, currently available HDR capturing sensors (e.g., smartphones) only improve the dynamic range by a few stops over conventional SDR. 

Associated SPS Event: IEEE ICIP 2022 Grand Challenge

The perceptual quality of images/videos in the context of video surveillance has a very significant impact on high-level tasks such as object detection, identification of abnormal events, visual tracking, to name a few. Despite the development of advanced video sensors with higher resolution, the quality of the acquired video is often affected by some distortions due to the environment, encoding and storage technologies, which can only be avoided by employing of intelligent post-processing solutions.

Face recognition in static images and video sequences captured in unconstrained recording conditions is one of the most widely studied topics in computer vision due to its extensive applications in surveillance, law enforcement, bio-metrics, marketing, and so forth. Recently, methodologies that achieve good performance have been presented in top-tier computer vision conferences (e.g., ICCV, CVPR, ECCV etc.) and great progress has been achieved in face recognition with deep learning-based methods.

Recent years have witnessed the great progress of the perception task such as image classification, object detection and pixel-wise semantic/instance segmentation. It is the right time to go one step further to infer the relations between the objects. Increasingly more efforts are devoted to relation prediction, such as the Visual Genome and Google Open Image challenge. There are mainly two differences between existing relation prediction works and PIC challenge.

This Challenge solicits contributions that demonstrate efficient algorithms for point cloud compression. Moreover, new rendering schemes, evaluation methodologies, as well as publicly accessible point cloud content are encouraged to be submitted, in addition to the proposed compression solutions.

To effectively prevent dengue fever outbreak, cleaning up the breeding sites of the mosquitos is essential. This proposal provides labeled data for the various types of containers, and aims to build an object detection model for possible breeding sites. This way the inspectors can pinpoint the containers which hold stagnant water by digital camera images or live video, and thus improve the effectiveness of inspection and breeding site elimination.

Habitat Challenge is an autonomous navigation challenge that aims to benchmark and accelerate progress in embodied AI. In its first iteration, Habitat Challenge 2019 is based on the PointGoal task defined in Anderson et. al. We will have two tracks for the PointGoal task:

1. RGB track: input modality for agent is RGB image.
2. RGBD track: input modalities for agent are RGB image and Depth.

The SUMO Challenge targets the development of algorithms for comprehensive understanding of 3D indoor scenes from 360° RGB-D panoramas. The target 3D models of indoor scenes include all visible layout elements and objects complete with pose, semantic information, and texture. Algorithms submitted are evaluated at 3 levels of complexity corresponding to 3 tracks of the challenge: oriented 3D bounding boxes, oriented 3D voxel grids, and oriented 3D meshes. SUMO Challenge results will be presented at the 2019 SUMO Challenge Workshop, at CVPR.

Outdoor scenes are often affected by fog, haze, rain, and smog. Poor visibility in the atmosphere is due to suspended particles. This challenge is meant to consolidate research efforts about single image recovering in adverse weather, especially hazy and rainy days. The challenge consists of two tracks: Hazy Image Recovering (HIR) and Rainy Image Recovering (RIR). In both tracks the researchers are required to recover sharp images from give degraded (hazy and rainy) inputs.

Automatic caption generation is the task of producing a natural-language utterance (usually a sentence) that describes the visual content of an image. Practical applications of automatic caption generation include leveraging descriptions for image indexing or retrieval, and helping those with visual impairments by transforming visual signals into information that can be communicated via text-to-speech technology. The CVPR 2019 Conceptual Captions Challenge is based on two separate test sets:

T1) a blind test set that participants do not have direct access to. 

Drones, or general UAVs, equipped with cameras have been fast deployed to a wide range of applications, including agricultural, aerial photography, fast delivery, and surveillance. Consequently, automatic understanding of visual data collected from these platforms become highly demanding, which brings computer vision to drones more and more closely. We are excited to present a large-scale benchmark with carefully annotated ground-truth for various important computer vision tasks, named VisDrone, to make vision meet drones.

The goal of the joint COCO and Mapillary Workshop is to study object recognition in the context of scene understanding. While both the COCO and Mapillary challenges look at the general problem of visual recognition, the underlying datasets and the specific tasks in the challenges probe different aspects of the problem.

As a continuous effort to push forward the research on video object segmentation tasks, we plan to host a second workshop with a challenge based on the YouTube-VOS dataset, targeting at more diversified problem settings, i.e., we plan to provide two challenge tracks in this workshop. The first track targets at semi-supervised video object segmentation, which is the same setting as in the first workshop. The second track will be a new task named video instance segmentation, which targets at automatically segmenting all object instances of pre-defined object categories from videos