The 2nd Learning from Imperfect Data (LID) Workshop

Weakly Supervised Learning for Real-World Computer Vision Applications

June 14th - 19th, 2020 Seattle, WA

Please feel free to contact us if you have any suggestions to improve our workshop!    lidworkshopcvpr@gmail.com

Schedule

Date June 14 Sunday 2020
(Pacific Time, SF Time)
Speaker Topic
8:30-8:40 Research Scientist at Stealth Mode AI Startup Shuai Zheng Opening Remark [slide][video][bilibili]
8:40-9:10 Professor at Johns Hopkins University Alan Yuille Invited talk 1: [slide][video][bilibili] You Only Annotate Once, or Never
9:10-9:40 CEO at Ariel AI Inc. / Associate Professor UCL Iasonas Kokkinos Invited talk 2: [slide][video][bilibili] Learning 3D object models from 2D images.
9:50-10:20 Director of Research at Mapillary Peter Kontschieder Invited talk 3: [slide][video][bilibili] Computer Vision with Less Supervision
10:30-11:00 Research Scientist at Google Boqing Gong Invited talk 4: [slide][video][bilibili] Towards Visual Recognition in the Wild: Long-Tailed Sources and Open Compound Targets
11:10-11:40 Staff Research Scientist at Facebook Research Zhicheng Yan Invited talk 5: [slide][video][bilibili] Decoupling Representation and Classifier for Long-Tailed Recognition
11:50-12:20 ETH Zurich Guolei Sun Oral 1: [slide][video][Q&A] The 1st Place of Track-1: Mining Cross-Image Semantics for Weakly Supervised Semantic Segmentation
14:00-14:30 Professor at University of California, Merced Ming-Hsuan Yang Invited talk 6: [slide][video][bilibili] Show, Match and Segment: Joint Weakly Supervised Learning of Semantic Matching and Object Co-segmentation
14:40-15:10 UCU & SoftServe Team Mariia Dobko Oral 2: [slide][video][Q&A] The 3rd Place of Track-1: NoPeopleAllowed: The 3 step approach to weakly supervised semantic segmentation
15:20-15:50 Intel Hao Zhao Oral 3: [slide][video][Q&A] The 1st Place of Track-2:Pointly supervised Scene Parsing with Uncertainty Mixture
16:00-16:30 Seoul National University Wonho Bae Oral 4: [slide][video][Q&A] The 1st Place of Track-3 & The 2nd Place of Track-1: Revisiting Class Activation Mapping for Learning from Imperfect Data
16:40-17:10 Beijing Jiaotong University & Mepro Team Chuangchuang Tan Oral 5: [slide][video][Q&A] The 2nd Place of Track-3: Dual Gradients Localization framework for Weakly Supervised Object Localization
17:20-17:50 Nanjing University Of Science and Technology & LEAP Group@PCA Lab Zhendong Wang Oral 6: [slide][video][bilibili][Q&A] The 3rd Place of Track-3: Weakly Supervised Object Localization
17:50-18:20 Professor at University of Adelaide Chunhua Shen Invited talk 7: [slide] Single shot instance segmentation
18:20-18:30 Assistant Professor at University of Technology Sydney Yunchao Wei Closing Remark [slide][video][bilibili]

Challenge results

Track1: Weakly-supervised Semantic Segmentation Challenge
1st Guolei Sun, Wenguan Wang, Luc Van Gool. ETH Zurich
2nd Wonho Bae*, Junhyug Noh*, Jinhwan Seo, and Gunhee Kim. Seoul National University, Vision & Learning Lab
3rd Mariia Dobko, Ostap Viniavskyi, Oles Dobosevych. The Machine Learning Lab at Ukrainian Catholic University, SoftServe
Track2: Weakly-supervised Scene Parsing Challenge
1st Hao Zhao, Ming Lu, Anbang Yao, Yiwen Guo, Yurong Chen, Li Zhang. Tsinghua University
Track3: Weakly-supervised Object Localization Challenge
1st Wonho Bae*, Junhyug Noh*, Jinhwan Seo, and Gunhee Kim. Seoul National University, Vision & Learning Lab
2nd Chuangchuang Tan 1*, Tao Ruan 1*, Guanghua Gu 2, Shikui Wei 1, Yao Zhao1. 1Beijing Jiaotong University, 2Yanshan University
3rd Zhendong Wang, Zhenyuan Chen, Chen Gong. Nanjing University Of Science and Technology, LEAP Group@PCA Lab

Papers

1st (Track1)

Mining Cross-Image Semantics for Weakly Supervised Semantic Segmentation (Best Paper Award)
Guolei Sun , Wenguan Wang , Luc Van Gool
The 2020 Learning from Imperfect Data (LID) Challenge - CVPR Workshops, 2020
[PDF] []

@article{sun2020lid,
     author = {Sun, Guolei and Wang, Wenguan and Van Gool, Luc},
     title = {Mining Cross-Image Semantics for Weakly Supervised Semantic Segmentation},
     journal = {The 2020 Learning from Imperfect Data (LID) Challenge - CVPR Workshops},
     year = {2020}
}

3rd (Track1)

NoPeopleAllowed: The Three-Step Approach to Weakly Supervised Semantic Segmentation
Mariia Dobko, Ostap Viniavskyi, Oles Dobosevych
The 2020 Learning from Imperfect Data (LID) Challenge - CVPR Workshops, 2020
[PDF] []

@article{dobko2020lid,
     author = {Dobko, Mariia and Viniavskyi, Ostap and Dobosevych, Oles},
     title = {NoPeopleAllowed: The Three-Step Approach to Weakly Supervised Semantic Segmentation},
     journal = {The 2020 Learning from Imperfect Data (LID) Challenge - CVPR Workshops},
     year = {2020}
}

1st (Track2)

Pointly-supervised Scene Parsing with Uncertainty Mixture
Hao Zhao, Ming Lu, Anbang Yao, Yiwen Guo, Yurong Chen, Li Zhang
The 2020 Learning from Imperfect Data (LID) Challenge - CVPR Workshops, 2020
[PDF] []

@article{zhao2020lid,
     author = {Zhao, Hao and Lu, Ming and Yao, Anbang and Guo, Yiwen and Chen, Yurong, Zhang, Li},
     title = {Pointly-supervised Scene Parsing with Uncertainty Mixture},
     journal = {The 2020 Learning from Imperfect Data (LID) Challenge - CVPR Workshops},
     year = {2020}
}

1st (Track3)

Revisiting Class Activation Mapping for Learning from Imperfect Data
Wonho Bae*, Junhyug Noh*, Jinhwan Seo, Gunhee Kim
The 2020 Learning from Imperfect Data (LID) Challenge - CVPR Workshops, 2020
[PDF] []

@article{bae2020lid,
     author = {Bae, Wonho and Noh, Junhyug and Seo, Jinhwan and Kim, Gunhee},
     title = {Revisiting Class Activation Mapping for Learning from Imperfect Data},
     journal = {The 2020 Learning from Imperfect Data (LID) Challenge - CVPR Workshops},
     year = {2020}
}

5th (Track3)

Object Localization with weakly supervised learning
Jun He, Huanqing Yan
The 2020 Learning from Imperfect Data (LID) Challenge - CVPR Workshops, 2020
[PDF] []

@article{he2020lid,
     author = {He, Jun and Yan, Huanqing},
     title = {Object Localization with weakly supervised learning},
     journal = {The 2020 Learning from Imperfect Data (LID) Challenge - CVPR Workshops},
     year = {2020}
}

Introduction

Weakly supervised learning refers to a variety of studies that attempt to address the challenging pattern recognition tasks by learning from weak or imperfect supervision. Supervised learning methods including Deep Convolutional Neural Networks (DCNNs) have significantly improved the performance in many problems in the field of computer vision, thanks to the rise of large-scale annotated data set and the advance in computing hardware. However, these supervised learning approaches are notorious "data hungry", which makes them are sometimes not practical in many real-world industrial applications. We are often facing the problem that we are not able to acquire enough amount of perfect annotations (e.g., object bounding boxes and pixel-wise masks) for reliable training models. To address this problem, many efforts in so-called weakly supervised learning approaches have been made to improve the DCNNs training to deviate from traditional paths of supervised learning using imperfect data. For instance, various approaches have proposed new loss functions or novel training schemes. Weakly supervised learning is a popular research direction in Computer Vision and Machine Learning communities, many research works have been devoted to related topics, leading to rapid growth of related publications in the top-tier conferences and journals such as CVPR, ICCV, ECCV, T-IP, and T-PAMI. We organize this workshop to investigate current ways of building industry level AI system relying on learning from imperfect data. We hope this workshop will attract attention and discussions from both industry and academic people.

Important Dates

Description Date
Challenge Begin 11:59PM Pacific Time Mar 22, 2020
Challenge Deadline 11:59PM Pacific Time June 8, 2020

People