Filters








14 Hits in 2.0 sec

Accuracy on the Line: On the Strong Correlation Between Out-of-Distribution and In-Distribution Generalization [article]

John Miller, Rohan Taori, Aditi Raghunathan, Shiori Sagawa, Pang Wei Koh, Vaishaal Shankar, Percy Liang, Yair Carmon, Ludwig Schmidt
2021 arXiv   pre-print
To complete the picture, we also investigate cases where the correlation is weaker, for instance some synthetic distribution shifts from CIFAR-10-C and the tissue classification dataset Camelyon17-WILDS  ...  out-of-distribution performance on variants of CIFAR-10 & ImageNet, a synthetic pose estimation task derived from YCB objects, satellite imagery classification in FMoW-WILDS, and wildlife classification in iWildCam-WILDS  ...  The iWildCam 2020 competition dataset.  ... 
arXiv:2107.04649v2 fatcat:ty44g72ypvdrzbhehftgpsdrja

Gradient Matching for Domain Generalization [article]

Yuge Shi, Jeffrey Seely, Philip H.S. Torr, N. Siddharth, Awni Hannun, Nicolas Usunier, Gabriel Synnaeve
2021 arXiv   pre-print
Our method produces competitive results on these datasets and surpasses all baselines on 4 of them.  ...  We demonstrate the efficacy of Fish on 6 datasets from the Wilds benchmark, which captures distribution shift across a diverse range of modalities.  ...  It outperforms all baseline on 4 datasets and achieves similar level of performance to the best method on the other 2 (AMAZON and IWILDCAM).  ... 
arXiv:2104.09937v3 fatcat:4obohplqwfd2fmgh7erk4nb4re

Detecting Errors and Estimating Accuracy on Unlabeled Data with Self-training Ensembles [article]

Jiefeng Chen, Frederick Liu, Besim Avci, Xi Wu, Yingyu Liang, Somesh Jha
2022 arXiv   pre-print
For example, on iWildCam, one instantiation reduces the estimation error for unsupervised accuracy estimation by at least 70% and improves the F1 score for error detection by at least 4.7% compared to  ...  For safe deployment, it is essential to estimate the accuracy of the pre-trained model on the test data.  ...  Acknowledgments The authors would like to thank Dr. Ankur Taly for his valuable comments.  ... 
arXiv:2106.15728v3 fatcat:4pogmew3abakrgsdnsqvoymbb4

Scaling biodiversity monitoring for the data age

Sara Beery
2021 XRDS Crossroads The ACM Magazine for Students  
Workshop at KDD 2019, https://sites.google.com/ corp/usc.edu/kdd19-dmaic-workshop/home 4 iWildCam 2021; https://sites.google.com/ view/fgvc8/competitions/iwildcam2021 5 GeoLifeCLEF 2020; https://www.imageclef  ...  From the point of view of a wildlife researcher, the plant images are bycatch that they don't have the capacity to label or curate from their existing datasets.  ... 
doi:10.1145/3466857 fatcat:4mn2tocl6bddrhnldcaa7ovr2q

Improving Baselines in the Wild [article]

Kazuki Irie, Imanol Schlag, Róbert Csordás, Jürgen Schmidhuber
2021
Our study focuses on two datasets: iWildCam and FMoW.  ...  difficult for iWildCam, (3) Minor changes in the training of hyper-parameters improve the baseline by a relatively large margin (mainly on FMoW), (4) There is a strong correlation between certain domains  ...  This work was partially supported by computational resources at the CSCS Swiss National Supercomputing Centre, project d115.  ... 
doi:10.48550/arxiv.2112.15550 fatcat:yvfg3nn77zfdjkpxiduckc5oae

WILDS: A Benchmark of in-the-Wild Distribution Shifts [article]

Pang Wei Koh, Shiori Sagawa, Henrik Marklund, Sang Michael Xie, Marvin Zhang, Akshay Balsubramani, Weihua Hu, Michihiro Yasunaga, Richard Lanas Phillips, Irena Gao, Tony Lee, Etienne David (+11 others)
2021 arXiv   pre-print
Despite their ubiquity in the real-world deployments, these distribution shifts are under-represented in the datasets widely used in the ML community today.  ...  Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild.  ...  acknowledge the support of DARPA under Nos.  ... 
arXiv:2012.07421v3 fatcat:bsohmukpszajxeadeo25oxmbs4

Extending the WILDS Benchmark for Unsupervised Adaptation [article]

Shiori Sagawa, Pang Wei Koh, Tony Lee, Irena Gao, Sang Michael Xie, Kendrick Shen, Ananya Kumar, Weihua Hu, Michihiro Yasunaga, Henrik Marklund, Sara Beery, Etienne David (+8 others)
2022 arXiv   pre-print
In this work, we present the WILDS 2.0 update, which extends 8 of the 10 datasets in the WILDS benchmark of distribution shifts to include curated unlabeled data that would be realistically obtainable  ...  On these datasets, we systematically benchmark state-of-the-art methods that leverage unlabeled data, including domain-invariant, self-training, and self-supervised methods, and show that their success  ...  The design of the WILDS benchmark was inspired by the Open Graph Benchmark (Hu et al., 2020b) , and we are grateful to the Open Graph Benchmark team for their advice and help in setting up our benchmark  ... 
arXiv:2112.05090v2 fatcat:77b2lkj7yvemteazktwxp4lnda

Adaptive Risk Minimization: Learning to Adapt to Domain Shift [article]

Marvin Zhang, Henrik Marklund, Nikita Dhawan, Abhishek Gupta, Sergey Levine, Chelsea Finn
2021 arXiv   pre-print
A fundamental assumption of most machine learning algorithms is that the training and test data are drawn from the same underlying distribution.  ...  In this work, we consider the problem setting of domain generalization, where the training data are structured into domains and there may be multiple test time shifts, corresponding to new domains or domain  ...  ARM-BN struggles on FMoW but performs well on the other datasets, in particular RxRx1. iWildCam Camelyon17 RxRx1 FMoW PovertyMap Method Acc Macro F1 Acc Acc WC Acc Avg Acc WC Pearson r Pearson r ERM 71.6  ... 
arXiv:2007.02931v4 fatcat:2gjsygbev5cmhdryhrwijvfeju

Conservation Tech Directory [article]

Carly Batist, Gracie Ermie
2021 figshare.com  
out the google form (link below) and we'll add it.https://forms.gle/qb9U8eFrsTWMGvwe6Suggested citation: Batist, C.H. & Ermi G. (2021).  ...  Feel free to reach out to them with any questions or suggestions regarding the directory!  ...  Open University Organization iWildCam annual Kaggle competition developing models to classify species and count individual animals across sequences in camera trap datasets https://github.com/  ... 
doi:10.6084/m9.figshare.15442200.v1 fatcat:psz6udw76nb7rdmp32teechvfi

Conservation Tech Directory [article]

Carly Batist, Gracie Ermi, Gracie Ermie
2021 figshare.com  
You can reach her on Twitter (@GracieErmie) or LinkedIn.Suggested citation: Batist, C.H. & Ermi G. (2021). Conservation technology directory.  ...  But please search the spreadsheet first to make sure the resource you're thinking of isn't already listed. Thanks!  ...  org/ The Open University citizen science images video education networking software Opportunity iWildCam annual Kaggle competition developing models to classify species and count individual  ... 
doi:10.6084/m9.figshare.15442200.v6 fatcat:rimzpsafq5foxf25n7ucgueh3a

Conservation Tech Directory [article]

Carly Batist, Gracie Ermi, Gracie Ermie
2021 figshare.com  
You can reach her on Twitter (@GracieErmie) or LinkedIn.Suggested citation: Batist, C.H. & Ermi G. (2021). Conservation technology directory.  ...  But please search the spreadsheet first to make sure the resource you're thinking of isn't already listed. Thanks!  ...  Open University citizen science images video education networking software Organization iWildCam https://github.com/visipedia /iwildcam_comp annual Kaggle competition developing models  ... 
doi:10.6084/m9.figshare.15442200.v4 fatcat:gzksivxqk5b45li4ltfubqj4fa

Conservation Tech Directory [article]

Carly Batist, Gracie Ermi, Gracie Ermie
2022 figshare.com  
You can reach her on Twitter (@GracieErmie) or LinkedIn.Suggested citation: Batist, C.H. & Ermi G. (2021). Conservation technology directory.  ...  But please search the spreadsheet first to make sure the resource you're thinking of isn't already listed. Thanks!  ...  large image, audio, or video datasets using computer vision light competitions annual workshop as part of ICML (2020 webinars recorded & online) term monitoring of biodiversity activity; deploying a prototype  ... 
doi:10.6084/m9.figshare.15442200.v13 fatcat:qgs2d2chsbdo5hflfzwu3p2rzq

Conservation Tech Directory [article]

Carly Batist, Gracie Ermi, Gracie Ermie
2022 figshare.com  
You can reach her on Twitter (@GracieErmie) or LinkedIn.Suggested citation: Batist, C.H. & Ermi G. (2021). Conservation technology directory.  ...  But please search the spreadsheet first to make sure the resource you're thinking of isn't already listed. Thanks!  ...  make things; gives step-by-step instruction guides for thousands of projects, including circuits & electronics; users can post their own guides to add to the platform iWildCam annual Kaggle competition  ... 
doi:10.6084/m9.figshare.15442200.v12 fatcat:b4bfumzexjec5lh4bdk3q35bia

INTERN: A New Learning Paradigm Towards General Vision [article]

Jing Shao, Siyu Chen, Yangguang Li, Kun Wang, Zhenfei Yin, Yinan He, Jianing Teng, Qinghong Sun, Mengya Gao, Jihao Liu, Gengshi Huang, Guanglu Song (+15 others)
2021
We evaluate our model on 26 well-known datasets that cover four categories of tasks in computer vision.  ...  Enormous waves of technological innovations over the past several years, marked by the advances in AI technologies, are profoundly reshaping the industry and the society.  ...  dataset creation.  ... 
doi:10.48550/arxiv.2111.08687 fatcat:lrjfbzax6jfrnpw4txnivh6cgy