{"type":"ScholarlyArticle","title":"Explaining Self-Supervised Image Representations with Visual Probing","authors":[{"name":"Dominika Basaj","slug":"dominika-basaj","given":"Dominika","family":"Basaj","affiliation":"Warsaw University of Technology"},{"name":"Witold Oleszkiewicz","slug":"witold-oleszkiewicz","given":"Witold","family":"Oleszkiewicz","affiliation":"Warsaw University of Technology"},{"name":"Igor Sieradzki","slug":"igor-sieradzki","given":"Igor","family":"Sieradzki","affiliation":"Faculty of Mathematics and Computer Science , Jagiellonian University"},{"name":"Michał Górszczak","slug":"micha-gorszczak","given":"Michał","family":"Górszczak","affiliation":"Faculty of Mathematics and Computer Science , Jagiellonian University"},{"name":"Barbara Rychalska","slug":"barbara-rychalska","given":"Barbara","family":"Rychalska","affiliation":"Warsaw University of Technology; Synerise"},{"name":"Tomasz Trzcinski","slug":"tomasz-trzcinski","given":"Tomasz","family":"Trzcinski","affiliation":"Warsaw University of Technology; Faculty of Mathematics and Computer Science , Jagiellonian University"},{"name":"Bartosz Zieliński","slug":"bartosz-zielinski","given":"Bartosz","family":"Zieliński","affiliation":"Faculty of Mathematics and Computer Science , Jagiellonian University"}],"abstract":"Recently introduced self-supervised methods for image representation learning provide on par or superior results to their fully supervised competitors, yet the corresponding efforts to explain the self-supervised approaches lag behind. Motivated by this observation, we introduce a novel visual probing framework for explaining the self-supervised models by leveraging probing tasks employed previously in natural language processing. The probing tasks require knowledge about semantic relationships between image parts. Hence, we propose a systematic approach to obtain analogs of natural language in vision, such as visual words, context, and taxonomy. We show the effectiveness and applicability of those analogs in the context of explaining self-supervised representations. Our key findings emphasize that relations between language and vision can serve as an effective yet intuitive tool for discovering how machine learning models work, independently of data modality. Our work opens a plethora of research pathways towards more explainable and transparent AI.","doi":"10.24963/ijcai.2021/82","url":"https://ijcai.org/proceedings/2021/82","pdf":"https://ijcai.org/proceedings/2021/0082.pdf","markdown":"https://ijcai.org/proceedings/2021/82.md","bibtex":"https://ijcai.org/proceedings/2021/bibtex/82","volume":{"slug":"2021","title":"Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2021","url":"https://ijcai.org/proceedings/2021/","year":2021,"location":"Montreal","dates":"19-27 August 2021","editors":"Prof. Zhi-Hua Zhou","isbn":"978-0-9992411-9-6","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2021"},"section":"Computer Vision I","track":"Main Track","number":82,"pages":{"first":592,"last":598},"keywords":["AI Ethics, Trust, Fairness: Explainability","Machine Learning: Unsupervised Learning","Computer Vision: Language and Vision"],"videos":[],"references":[{"position":1,"raw":"[Achanta et al., 2012] R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua, and S. Süsstrunk. Slic superpixels com- pared to state-of-the-art superpixel methods. IEEE Trans- actions on Pattern Analysis and Machine Intelligence, 34(11), 2012.","title":"Slic superpixels compared to state-of-the-art superpixel methods","authors":"R Achanta, A Shaji, K Smith, A Lucchi, P Fua, S Süsstrunk","year":"2012","venue":"IEEE Transactions on Pattern Analysis and Machine Intelligence","doi":null,"url":null,"internal_url":null},{"position":2,"raw":"[Adebayo et al., 2018] J. Adebayo, J. Gilmer, M. Muelly, I. Goodfellow, M. Hardt, and B. Kim. Sanity checks for saliency maps. In NeurIPS, 2018.","title":"Sanity checks for saliency maps","authors":"J Adebayo, J Gilmer, M Muelly, I Goodfellow, M Hardt, B Kim","year":"2018","venue":"NeurIPS","doi":null,"url":null,"internal_url":null},{"position":3,"raw":"[Alain and Bengio, 2017] G. Alain and Y. Bengio. Under- standing intermediate layers using linear classifier probes. In ICLR Workshop, 2017.","title":"Understanding intermediate layers using linear classifier probes","authors":"G Alain, Y Bengio","year":"2017","venue":"ICLR Workshop","doi":null,"url":null,"internal_url":null},{"position":4,"raw":"[Antol et al., 2015] S. Antol, A. Agrawal, J. Lu, M. Mitchell, D. Batra, C. L. Zitnick, and D. Parikh. VQA: Visual Ques- tion Answering. In ICCV, 2015.","title":"VQA: Visual Question Answering","authors":"S Antol, A Agrawal, J Lu, M Mitchell, D Batra, C L Zitnick, D Parikh","year":"2015","venue":"ICCV","doi":null,"url":null,"internal_url":null},{"position":5,"raw":"[Carion et al., 2020] N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko. End-to-end object detection with transformers. In ECCV, 2020.","title":"End-to-end object detection with transformers","authors":"N Carion, F Massa, G Synnaeve, N Usunier, A Kirillov, S Zagoruyko","year":"2020","venue":"ECCV","doi":null,"url":null,"internal_url":null},{"position":6,"raw":"[Caron et al., 2020] M. Caron, I. Misra, J. Mairal, P. Goyal, P. Bojanowski, and A. Joulin. Unsupervised learning of visual features by contrasting cluster assignments. arXiv preprint 2006.09882, 2020.","title":"Unsupervised learning of visual features by contrasting cluster assignments","authors":"M Caron, I Misra, J Mairal, P Goyal, P Bojanowski, A Joulin","year":"2020","venue":null,"doi":null,"url":null,"internal_url":null},{"position":7,"raw":"[Chen et al., 2020a] T. Chen, S. Kornblith, M. Norouzi, and G. Hinton. A simple framework for contrastive learning of visual representations. In ICML, 2020.","title":"A simple framework for contrastive learning of visual representations","authors":"Chen","year":"2020","venue":"ICML","doi":null,"url":null,"internal_url":null},{"position":8,"raw":"[Chen et al., 2020b] T. Chen, S. Kornblith, K. Swersky, M. Norouzi, and G. Hinton. Big self-supervised models are strong semi-supervised learners. In NeurIPS, 2020.","title":"Big self-supervised models are strong semi-supervised learners","authors":"Chen","year":"2020","venue":"NeurIPS","doi":null,"url":null,"internal_url":null},{"position":9,"raw":"[Conneau et al., 2018] A. Conneau, G. Kruszewski, G. Lam- ple, L. Barrault, and M. Baroni. What you can cram into a single $&!#* vector: Probing sentence embeddings for linguistic properties. In ACL, 2018.","title":"What you can cram into a single $&!#* vector: Probing sentence embeddings for linguistic properties","authors":"A Conneau, G Kruszewski, G Lample, L Barrault, M Baroni","year":"2018","venue":"ACL","doi":null,"url":null,"internal_url":null},{"position":10,"raw":"[Deng et al., 2009] J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei. ImageNet: A Large-Scale Hierar- chical Image Database. In CVPR, 2009.","title":"ImageNet: A Large-Scale Hierarchical Image Database","authors":"J Deng, W Dong, R Socher, L.-J Li, K Li, L Fei-Fei","year":"2009","venue":"CVPR","doi":null,"url":null,"internal_url":null},{"position":11,"raw":"[Dosovitskiy et al., 2020] A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint 2010.11929, 2020.","title":"An image is worth 16x16 words: Transformers for image recognition at scale","authors":"A Dosovitskiy, L Beyer, A Kolesnikov, D Weissenborn, X Zhai, T Unterthiner, M Dehghani, M Minderer, G Heigold, S Gelly, J Uszkoreit, N Houlsby","year":"2020","venue":null,"doi":null,"url":null,"internal_url":null},{"position":12,"raw":"[Eichler et al., 2019] M. Eichler, G. G. S ¸ahin, and Ir Gurevych. LINSPECTOR WEB: A multilingual probing suite for word representations. In EMNLP- IJCNLP: System Demonstrations, 2019.","title":"LINSPECTOR WEB: A multilingual probing suite for word representations","authors":"M Eichler, G G, Ir Gurevych","year":"2019","venue":"EMNLP-IJCNLP: System Demonstrations","doi":null,"url":null,"internal_url":null},{"position":13,"raw":"[Geirhos et al., 2020] R. Geirhos, K. Narayanappa, B. Mitzkus, M. Bethge, F. A. Wichmann, and W. Brendel. On the surprising similarities between supervised and self-supervised models. In NeurIPS Workshop, 2020.","title":"On the surprising similarities between supervised and self-supervised models","authors":"R Geirhos, K Narayanappa, B Mitzkus, M Bethge, F A Wichmann, W Brendel","year":"2020","venue":"NeurIPS Workshop","doi":null,"url":null,"internal_url":null},{"position":14,"raw":"[Ghorbani et al., 2019] A. Ghorbani, James Wexler, J. Zou, and Been Kim. Towards automatic concept-based expla- nations. In NeurIPS, 2019.","title":"Towards automatic concept-based explanations","authors":"A Ghorbani, James Wexler, J Zou, Been Kim","year":"2019","venue":"NeurIPS","doi":null,"url":null,"internal_url":null},{"position":15,"raw":"[Grill et al., 2020] J.-B. Grill, F. Strub, F. Altché, C. Tal- lec, P. H. Richemond, E. Buchatskaya, C. Doersch, B. A. Pires, Z. D. Guo, M. G. Azar, B. Piot, K. Kavukcuoglu, R. Munos, and M. Valko. Bootstrap your own latent: A new approach to self-supervised learning. arXiv preprint 2006.07733, 2020.","title":"Bootstrap your own latent: A new approach to self-supervised learning","authors":"J.-B Grill, F Strub, F Altché, C Tallec, P H Richemond, E Buchatskaya, C Doersch, B A Pires, Z D Guo, M G Azar, B Piot, K Kavukcuoglu, R Munos, M Valko","year":"2020","venue":null,"doi":null,"url":null,"internal_url":null},{"position":16,"raw":"[Hadsell et al., 2006] R. Hadsell, S. Chopra, and Y. LeCun. Dimensionality reduction by learning an invariant map- ping. In CVPR, 2006.","title":"Dimensionality reduction by learning an invariant mapping","authors":"R Hadsell, S Chopra, Y Lecun","year":"2006","venue":"CVPR","doi":null,"url":null,"internal_url":null},{"position":17,"raw":"[He et al., 2019] Kaiming He, Haoqi Fan, Yuxin Wu, Sain- ing Xie, and Ross B. Girshick. Momentum contrast for un- supervised visual representation learning. arXiv preprint 1911.05722, 2019.","title":"Momentum contrast for unsupervised visual representation learning","authors":"Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, Ross B Girshick","year":"2019","venue":null,"doi":null,"url":null,"internal_url":null},{"position":18,"raw":"[Hendrycks et al., 2019] D. Hendrycks, M. Mazeika, S. Ka- davath, and D. Song. Using self-supervised learning can improve model robustness and uncertainty. In NeurIPS, 2019.","title":"Using self-supervised learning can improve model robustness and uncertainty","authors":"D Hendrycks, M Mazeika, S Kadavath, D Song","year":"2019","venue":"NeurIPS","doi":null,"url":null,"internal_url":null},{"position":19,"raw":"[Huang and Li, 2020] Zixuan Huang and Yin Li. Inter- pretable and accurate fine-grained recognition via region grouping. In CVPR, 2020.","title":"Interpretable and accurate fine-grained recognition via region grouping","authors":"Li ; Huang, Zixuan Huang, Yin Li","year":"2020","venue":"CVPR","doi":null,"url":null,"internal_url":null},{"position":20,"raw":"[Kim et al., 2018] B. Kim, M. Wattenberg, J. Gilmer, C. Cai, J. Wexler, F. Viegas, and R. Sayres. Interpretability be- yond feature attribution: Quantitative testing with concept activation vectors (tcav). In ICML, 2018.","title":"Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (tcav)","authors":"B Kim, M Wattenberg, J Gilmer, C Cai, J Wexler, F Viegas, R Sayres","year":"2018","venue":null,"doi":null,"url":null,"internal_url":null},{"position":21,"raw":"[Krizhevsky et al., 2012] A. Krizhevsky, I. Sutskever, and G. Hinton. Imagenet classification with deep convolu- tional neural networks. In NeurIPS, 2012.","title":"Imagenet classification with deep convolutional neural networks","authors":"A Krizhevsky, I Sutskever, G Hinton","year":"2012","venue":"NeurIPS","doi":null,"url":null,"internal_url":null},{"position":22,"raw":"[Kumar and Talukdar, 2020] Sawan Kumar and Partha Talukdar. NILE : Natural language inference with faithful natural language explanations. In ACL, 2020.","title":"NILE : Natural language inference with faithful natural language explanations","authors":"Talukdar Kumar, Sawan Kumar, Partha Talukdar","year":"2020","venue":"ACL","doi":null,"url":null,"internal_url":null},{"position":23,"raw":"[Leung and Malik, 2001] T. Leung and J. Malik. Represent- ing and recognizing the visual appearance of materials us- ing three-dimensional textons. In International journal of computer vision. Springer, 2001.","title":"Representing and recognizing the visual appearance of materials using three-dimensional textons","authors":"T Leung, J Malik","year":"2001","venue":"International journal of computer vision","doi":null,"url":null,"internal_url":null},{"position":24,"raw":"[Simonyan et al., 2014] K. Simonyan, A. Vedaldi, and A. Zisserman. Deep inside convolutional networks: Vi- sualising image classification models and saliency maps. In ICLR Workshop, 2014.","title":"Deep inside convolutional networks: Visualising image classification models and saliency maps","authors":"K Simonyan, A Vedaldi, A Zisserman","year":"2014","venue":"ICLR Workshop","doi":null,"url":null,"internal_url":null},{"position":25,"raw":"[Sivic and Zisserman, 2006] Josef Sivic and Andrew Zisser- man. Video google: Efficient visual search of videos. In Toward Category-Level Object Recognition. Springer, 2006.","title":"Video google: Efficient visual search of videos","authors":"Zisserman Sivic, Josef Sivic, Andrew Zisserman","year":"2006","venue":"Toward Category-Level Object Recognition","doi":null,"url":null,"internal_url":null},{"position":26,"raw":"[Szegedy et al., 2015] C. Szegedy, W. Liu, Y. Jia, P. Ser- manet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, and A. Rabinovich. Going deeper with convolutions. In CVPR, 2015.","title":"Going deeper with convolutions","authors":"C Szegedy, W Liu, Y Jia, P Sermanet, S Reed, D Anguelov, D Erhan, V Vanhoucke, A Rabinovich","year":"2015","venue":"CVPR","doi":null,"url":null,"internal_url":null},{"position":27,"raw":"[Vulić et al., 2020] I. Vulić, E. M. Ponti, R. Litschko, G. Glavaš, and A. Korhonen. Probing pretrained lan- guage models for lexical semantics. arXiv preprint arXiv:2010.05731, 2020.","title":"Probing pretrained language models for lexical semantics","authors":"Vulić","year":"2020","venue":null,"doi":null,"url":null,"internal_url":null},{"position":28,"raw":"[Zhang and Zhu, 2018] Quan-shi Zhang and Song-chun Zhu. Visual interpretability for deep learning: A survey. In Frontiers of Information Technology & Electronic En- gineering, 2018.","title":"Quan-shi Zhang and Song-chun Zhu. Visual interpretability for deep learning: A survey","authors":"Zhu Zhang","year":"2018","venue":"Frontiers of Information Technology & Electronic Engineering","doi":null,"url":null,"internal_url":null}],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}