CVPR 2021 競(jìng)賽匯總

Neural Architecture Search

1st lightweight NAS challenge and moving beyond

https://www.cvpr21-nas.com/competition

早期的NAS方法通過(guò)將每個(gè)神經(jīng)網(wǎng)絡(luò)在訓(xùn)練數(shù)據(jù)上都訓(xùn)練到收斂，然后評(píng)估其效果，需要耗費(fèi)大量的算力資源。

Track1:Supernet Track

賽道一為超網(wǎng)絡(luò)賽道，旨在解決OneshotNAS的一致性問(wèn)題；

Track2: Performance Prediction Track

賽道二為模型性能預(yù)測(cè)賽道，旨在不做任何訓(xùn)練的情況，準(zhǔn)確的預(yù)測(cè)任意模型結(jié)構(gòu)在特定評(píng)測(cè)集的性能。

Track3: Dataset-Agnostic Track

賽道三鼓勵(lì)參與者提交與數(shù)據(jù)無(wú)關(guān)，但能夠在完全未知的數(shù)據(jù)集上提供優(yōu)秀結(jié)果的NAS算法。

JackRabbot Social Grouping and Activity Dataset and Benchmark

2nd Workshop on Visual Perception for Navigation in Human Environments

https://jrdb.stanford.edu/workshops/jrdb-cvpr21

除了JRDB上現(xiàn)有的四個(gè)基準(zhǔn)和挑戰(zhàn)（即2D-3D人員檢測(cè)和跟蹤挑戰(zhàn)）之外，在本研討會(huì)中，我們使用新的注釋來(lái)組織兩個(gè)新的挑戰(zhàn)：

• 人類社會(huì)群體檢測(cè)
• 個(gè)人動(dòng)作檢測(cè)和社交活動(dòng)識(shí)別

NTIRE 2021 challenges

New Trends in Image Restoration and Enhancement workshop and challenges on image and video processing

https://data.vision.ee.ethz.ch/cvl/ntire21/

NTIRE Image challenges

• Nonhomogeneous Dehazing
• Defocus Deblurring using Dual-pixel
• Depth Guided Image Relighting: Track 1 One-to-One relighting
• Depth Guided Image Relighting: Track 2 Any-to-Any relighting
• Perceptual Image Quality Assessment
• Image Deblurring: Track 1 Low Resolution
• Image Deblurring: Track 2 JPEG Artifacts
• Multi-Modal Aerial View Imagery Classification: Track 1 (SAR)
• Multi-Modal Aerial View Imagery Classification: Track 2 (SAR+EO)
• Learning the Super-Resolution Space

NTIRE video/multi-frame challenges

• Quality enhancement of heavily compressed videos: Track 1 Fixed QP, Fidelity
• Quality enhancement of heavily compressed videos: Track 2 Fixed QP, Perceptual
• Quality enhancement of heavily compressed videos: Track 3 Fixed bit-rate, Fidelity
• Video Super-Resolution: Track 1 Spatial started!
• Video Super-Resolution: Track 2 Spatio-Temporal
• Burst Super-Resolution: Track 1 Synthetic
• Burst Super-Resolution: Track 2 Real
• High Dynamic Range (HDR): Track 1 Single frame
• High Dynamic Range (HDR): Track 2 Multiple frames

Mobile AI 2021 challenges

• Learned ISP (MediaTek Dimensity APU platform)
• Image Denoising (Samsung Exynos Mali GPU platform)
• HDR Image Processing (Huawei Kirin Da Vinci NPU platform)
• Image Super-Resolution (Synaptics Dolphin NPU platform)
• Video Super-Resolution (OPPO Snapdragon Adreno GPU platform)
• Depth Estimation (Raspberry Pi 4 platform)
• Camera Scene Detection (Apple Bionic platform)

SHApe Recovery from Partial Textured 3D Scans

該研討會(huì)的目的是推廣在3D掃描處理中同時(shí)利用形狀和紋理的概念，并特別注意從部分和嘈雜數(shù)據(jù)中恢復(fù)的特定任務(wù)。

https://cvi2.uni.lu/sharp2021/

Recovery of Human Body Scans

Recovery of Generic Object Scans

Recovery of Feature Edges in 3D Object Scans

LOVEU: LOng-form VidEo Understanding

https://sites.google.com/view/loveucvpr21/challenge

VizWiz Grand Challenge Workshop

https://vizwiz.org/workshops/2021-workshop/

Task: Image Captioning

Given an image, the task is to predict an accurate caption.

Task: Predict Answer to a Visual Question

Given an image and question about it, the task is to predict an accurate answer.

Task: Predict Answerability of a Visual Question

Given an image and question about it, the task is to predict if the visual question cannot be answered (with a confidence score in that prediction).

Bridging the Gap between Computational Photography and Visual Recognition

http://cvpr2021.ug2challenge.org/

TRACK 1: OBJECT DETECTION IN POOR VISIBILITY ENVIRONMENTS

TRACK 2: ACTION RECOGNITION FROM DARK VIDEOS

4th Workshop and Challenge on Learned Image Compression

image compression track

images need to be compressed to 0.075 bpp, 0.15 bpp, and 0.3 bpp (bits per pixel).

video compression track

short video clips need to be compressed to around 1 Mbit/s.

perceptual metric track

human preferences on pairs of images will have to be predicted. The image pairs will come from the decoders submitted to the image compression track.

5th AI City Challenge

https://www.aicitychallenge.org/

Challenge Track 1: Multi-Class Multi-Movement Vehicle Counting Using IoT Devices

Participating teams will count four-wheel vehicles and freight trucks that follow pre-defined movements from multiple camera scenes.

Challenge Track 2: City-Scale Multi-Camera Vehicle Re-Identification

Participating teams will perform vehicle re-identification based on vehicle crops from multiple cameras placed at multiple intersections.

Challenge Track 3: City-Scale Multi-Camera Vehicle Tracking

Participating teams will track vehicles across multiple cameras both at a single intersection and across multiple intersections spread out across a city.

Challenge Track 4: Traffic Anomaly Detection

Participating teams will submit all anomalies detected in the test data, including car crashes, stalled vehicles based on video feeds from multiple cameras at intersections and along highways.

Challenge Track 5: Natural Language-Based Vehicle Retrieval

Natural language (NL) description offers another useful way to specify vehicle track queries.

Large-scale Video Object Segmentation Challenge

https://youtube-vos.org/challenge/2021/

Our workshop has three challenges for different video segmentation tasks including semi-supervised video object segmentation, video instance segmentation and referring video object segmentation.

Track 1: Video Object Segmentation

Track 2: Video Instance Segmentation

Track 3: Referring Video Object Segmentation

Looking at People Large Scale Signer Independent Isolated SLR

http://chalearnlap.cvc.uab.es/challenge/43/description/

We are organizing a challenge on isolated sign language recognition from signer-independent non-controlled RGB-D data involving a large number of sign categories (>200).

RGB Competition Track

RGB+D Competition Track

3rd ScanNet Indoor Scene Understanding Challenge

http://www.scan-net.org/cvpr2021workshop/

International Challenge on Activity Recognition (ActivityNet)

http://activity-net.org/challenges/2021/

n this installment of the challenge, we will host seven guest tasks (tentative) focusing on different aspects of the activity recognition problem, especially expanding from online consumer video challenges to challenges on surveillance and first-person video.

Agriculture-Vision: Challenges & Opportunities for Computer Vision in Agriculture

https://www.agriculture-vision.com/

The 2nd Agriculture-Vision Prize Challenge aims to encourage research in developing novel and effective algorithms for agricultural pattern recognition from aerial images. Submissions will be evaluated and ranked by model performance.This year, we will be hosting two challenge tracks: supervised track and semi-supervised track. The top three performing submissions will receive prize rewards and presentation opportunities at our workshop.

Built Environment for the Design, Construction, and Operation of Buildings

https://cv4aec.github.io/

Semantic and Instance Segmentation of building elements

Object Attribute Prediction of building elements

Learning from Limited or Imperfect Data

https://l2id.github.io/

Learning from limited or imperfect data (L^2ID) refers to a variety of studies that attempt to address challenging pattern recognition tasks by learning from limited, weak, or noisy supervision.

Open World Vision

http://www.cs.cmu.edu/~shuk/open-world-vision.html#competition

Open-set image classification requires a model to distinguish novel, anomalous and semantically unknown (e.g., open-set) test-time examples.

Adversarial Machine Learning in Real-World Computer Vision Systems and Online Challenges

https://aisecure-workshop.github.io/amlcvpr2021/

Adversarial Attacks on ML Defense Models

Unrestricted Adversarial Attacks on ImageNet

Continual Learning in Computer Vision

https://eval.ai/web/challenges/challenge-page/829/overview

Robust Video Scene Understanding: Tracking and Video Segmentation

https://eval.vision.rwth-aachen.de/rvsu-workshop21/

EarthVision: Large Scale Computer Vision for Remote Sensing Imagery

http://www.classic.grss-ieee.org/earthvision2021/challenge.html

DynamicEarthNet Challenge

FloodNet Challenge

Image Matching: Local Features & Beyond

https://image-matching-workshop.github.io/

Chart Question Answering Workshop

https://cqaw.github.io/

The CQA challenge includes 3 levels of perception: from low-level visualization building blocks to semantic reasoning that requires text extraction.

2nd. Thermal Image Super-Resolution Challenge

https://pbvs-workshop.github.io/challenge.html

The Eight Workshop on Fine-Grained Visual Categorization

https://sites.google.com/view/fgvc8

• GeoLifeCLEF2021
• Semi-iNat2021
• iNatChallenge2021
• iMet2021
• iMat-Fashion2021
• Hotel-ID2021
• HerbariumChallenge2021
• iWildCam2021
• Plant Pathology Challenge 2021

GAZE 2021 Challenges

The GAZE 2021 Challenges are hosted on Codalab, and can be found at:

• ETH-XGaze Challenge: https://competitions.codalab.org/competitions/28930
• EVE Challenge: https://competitions.codalab.org/competitions/28954

Autonomous Navigation in Unconstrained Environments

http://cvit.iiit.ac.in/autonue2021/challenge/

• Challenges for domain adaptation with varying levels of supervision.
• Challenges for semantic segmentation.

其他鏈接

由于很多競(jìng)賽還在更新，完整競(jìng)賽參考CVPR官網(wǎng)：

• http://cvpr2021.thecvf.com/workshops-schedule

• https://github.com/skrish13/ml-contests-conf