Paced-curriculum distillation with prediction and label uncertainty for image segmentation

Springer Science and Business Media LLC - Tập 18 - Trang 1875-1883 - 2023
Mobarakol Islam1, Lalithkumar Seenivasan2, S. P. Sharan3, V. K. Viekash4, Bhavesh Gupta5, Ben Glocker1, Hongliang Ren2,6
1Department of Computing, Imperial College London, London, UK
2Department of Biomedical Engineering, National University of Singapore, Singapore, Singapore
3Department of Electronics and Communication Engineering, National Institute of Technology Tiruchirappalli, Tiruchirappalli, India
4Department of Instrumentation and Control Engineering, National Institute of Technology, Tiruchirappalli, Tiruchirappalli, India
5Department of Mechanical Engineering, Indian Institute of Technology Delhi, India
6Department of Electrical Engineering and Shun Hing Institute of Advanced Engineering, The Chinese University of Hong Kong, Hong Kong, China

Tóm tắt

In curriculum learning, the idea is to train on easier samples first and gradually increase the difficulty, while in self-paced learning, a pacing function defines the speed to adapt the training progress. While both methods heavily rely on the ability to score the difficulty of data samples, an optimal scoring function is still under exploration. Distillation is a knowledge transfer approach where a teacher network guides a student network by feeding a sequence of random samples. We argue that guiding student networks with an efficient curriculum strategy can improve model generalization and robustness. For this purpose, we design an uncertainty-based paced curriculum learning in self-distillation for medical image segmentation. We fuse the prediction uncertainty and annotation boundary uncertainty to develop a novel paced-curriculum distillation (P-CD). We utilize the teacher model to obtain prediction uncertainty and spatially varying label smoothing with Gaussian kernel to generate segmentation boundary uncertainty from the annotation. We also investigate the robustness of our method by applying various types and severity of image perturbation and corruption. The proposed technique is validated on two medical datasets of breast ultrasound image segmentation and robot-assisted surgical scene segmentation and achieved significantly better performance in terms of segmentation and robustness. P-CD improves the performance and obtains better generalization and robustness over the dataset shift. While curriculum learning requires extensive tuning of hyper-parameters for pacing function, the level of performance improvement suppresses this limitation.

Tài liệu tham khảo

Bengio Y, Louradour J, Collobert R, Weston J (2009) Curriculum learning. In: Proceedings of the 26th annual international conference on machine learning, pp 41– 48 Sinha S, Garg A, Larochelle H (2020) Curriculum by smoothing. Adv Neural Inf Process Syst 33:21653–21664 Kumar M, Packer B, Koller D (2010) Self-paced learning for latent variable models. Adv Neural Inf Process Syst 23:1189–1197 Jiang L, Meng D, Zhao Q, Shan S, Hauptmann AG ( 2015) Self-paced curriculum learning. In: Twenty-Ninth AAAI Conference on Artificial Intelligence Zhao H, Sun X, Dong J, Dong Z, Li Q (2021) Knowledge distillation via instance-level sequence learning. Knowl-Based Syst 233:107519 Wei J, Suriawinata A, Ren B, Liu X, Lisovsky M, Vaickus L, Brown C, Baker M, Nasir-Moin M, Tomita N, Torresani L, Wei J, Hassanpour S ( 2021) Learn like a pathologist: Curriculum learning by annotator agreement for histopathology image classification. In: Proceedings of the IEEE/cvf winter conference on applications of computer vision (WACV), pp 2473– 2483 Shi W, Song Y, Zhou H, Li B, Li L (2021) Follow your path: a progressive method for knowledge distillation. Joint European conference on machine learning and knowledge discovery in databases. Springer, Berlin, pp 596–611 Ye D, Chen G, Zhang W, Chen S, Yuan B, Liu B, Chen J, Liu Z, Qiu F, Yu H et al (2020) Towards playing full MOBA games with deep reinforcement learning. Adv Neural Inf Process Syst 33:621–632 Zhu Q, Chen X, Wu P, Liu J, Zhao D (2021) Combining curriculum learning and knowledge distillation for dialogue generation. In: Findings of the Association for Computational Linguistics: EMNLP, pp 1284– 1295 Xiang L, Ding G, Han J (2020) Learning from multiple experts: self-paced knowledge distillation for long-tailed classification. European conference on computer vision. Springer, Berlin, pp 247–263 Wang Y, Gan W, Yang J, Wu W, Yan J ( 2019) Dynamic curriculum learning for imbalanced data classification. In: Proceedings of the IEEE/CVF international conference on computer Vision, pp 5017– 5026 Chang H-S, Learned-Miller E, McCallum A (2017) Active bias: training more accurate neural networks by emphasizing high variance samples. Adv Neu Inform Proc Syst. 30 Zhou Y, Yang B, Wong DF, Wan Y, Chao LS ( 2020) Uncertainty-aware curriculum learning for neural machine translation. In: Proceedings of the 58th Annual Meeting of the association for computational linguistics, pp 6934– 6944 Platanios EA, Stretcu O, Neubig G, Poczos B, Mitchell TM ( 2019) Competence-based curriculum learning for neural machine translation. In: International conference of the north american chapter of the association for computational linguistics: Human Language Technologies, pp 1162– 1172 Zhang Y, Abbeel P, Pinto L (2020) Automatic curriculum learning through value disagreement. Adv Neural Inf Process Syst 33:7648–7659 Lee HJ, Kim JU, Lee S, Kim HG, Ro YM ( 2020) Structure boundary preserving segmentation for medical image with ambiguous boundary. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp 4817– 4826 Islam M, Glocker B ( 2021) Spatially varying label smoothing: Capturing uncertainty from expert annotations. In: international conference on information processing in medical imaging, pp 677– 688 . Springer Tang P, Yang P, Nie D, Wu X, Zhou J, Wang Y (2022) Unified medical image segmentation by learning from uncertainty in an end-to-end manner. Knowl-Based Syst 241:108215 Allan M, Kondo S, Bodenstedt S, Leger S, Kadkhodamohammadi R, Luengo I, Fuentes F, Flouty E, Mohammed AK, Pedersen M, Kori AK, Varghese A, Krishnamurthi G, Rauber D, Mendel R, Palm C, Bano S, Saibro G, Shih C, Chiang H, Zhuang J, Yang J, Iglovikov V, Dobrenkii A, Reddiboina M, Reddy A, Liu X, Gao C, Unberath M, Azizian M, Stoyanov D, Maier-Hein L, Speidel S (2020) 2018 robotic scene segmentation challenge. arXiv preprint arXiv:2001.11190 Al-Dhabyani W, Gomaa M, Khaled H, Fahmy A (2020) Dataset of breast ultrasound images. Data Brief 28:104863 Hinton G, Vinyals O, Dean J (2015) Distilling the knowledge in a neural network. arXiv preprint. 2(7) Kim K, Ji B, Yoon D, Hwang S ( 2021) Self-knowledge distillation with progressive refinement of targets. In: Proceedings of the IEEE/CVF international conference on computer vision, pp 6567– 6576 Guo C, Pleiss G, Sun Y, Weinberger KQ ( 2017) On calibration of modern neural networks. In: International conference on machine learning, pp 1321– 1330 . PMLR Ronneberger O, Fischer P, Brox T (2015) U-net: Convolutional networks for biomedical image segmentation. International conference on medical image computing and computer-assisted intervention. Springer, Berlin, pp 234–241 Chaurasia A, Culurciello E (2017) Linknet: Exploiting encoder representations for efficient semantic segmentation. In: 2017 IEEE visual communications and image processing (VCIP), pp 1– 4 . IEEE Chen L-C, Zhu Y, Papandreou G, Schroff F, Adam H ( 2018) Encoder-decoder with atrous separable convolution for semantic image segmentation. In: Proceedings of the european conference on computer vision (ECCV), pp 801– 818 Gong T, Zhao Q, Meng D, Xu Z (2016) Why curriculum learning & self-paced learning work in big/noisy data: a theoretical perspective. Big Data Inform Anal 1(1):111 Wu X, Dyer E, Neyshabur B (2020) When do curricula work? arXiv preprint arXiv:2012.03107 Hendrycks D, Dietterich T (2019) Benchmarking neural network robustness to common corruptions and perturbations. arXiv preprint arXiv:1903.12261