self training with noisy student improves imagenet classification

Finally, we iterate the algorithm a few times by treating the student as a teacher to generate new pseudo labels and train a new student. The pseudo labels can be soft (a continuous distribution) or hard (a one-hot distribution). all 12, Image Classification Self-training See We then train a larger EfficientNet as a student model on the combination of labeled and pseudo labeled images. Qizhe Xie, Eduard Hovy, Minh-Thang Luong, Quoc V. Le. Med. We thank the Google Brain team, Zihang Dai, Jeff Dean, Hieu Pham, Colin Raffel, Ilya Sutskever and Mingxing Tan for insightful discussions, Cihang Xie for robustness evaluation, Guokun Lai, Jiquan Ngiam, Jiateng Xie and Adams Wei Yu for feedbacks on the draft, Yanping Huang and Sameer Kumar for improving TPU implementation, Ekin Dogus Cubuk and Barret Zoph for help with RandAugment, Yanan Bao, Zheyun Feng and Daiyi Peng for help with the JFT dataset, Olga Wichrowska and Ola Spyra for help with infrastructure. Self-training with Noisy Student improves ImageNet classification Original paper: https://arxiv.org/pdf/1911.04252.pdf Authors: Qizhe Xie, Eduard Hovy, Minh-Thang Luong, Quoc V. Le HOYA012 Introduction EfficientNet ImageNet SOTA EfficientNet In contrast, changing architectures or training with weakly labeled data give modest gains in accuracy from 4.7% to 16.6%. The mapping from the 200 classes to the original ImageNet classes are available online.222https://github.com/hendrycks/natural-adv-examples/blob/master/eval.py. This material is presented to ensure timely dissemination of scholarly and technical work. Self-Training achieved the state-of-the-art in ImageNet classification within the framework of Noisy Student [1]. However, during the learning of the student, we inject noise such as dropout, stochastic depth and data augmentation via RandAugment to the student so that the student generalizes better than the teacher. Conclusion, Abstract , ImageNet , web-scale extra labeled images weakly labeled Instagram images weakly-supervised learning . Here we study how to effectively use out-of-domain data. w Summary of key results compared to previous state-of-the-art models. To date (2020) we will introduce "Noisy Student Training", which is a state-of-the-art model.The idea is to extend self-training and Distillation, a paper that shows that by adding three noises and distilling multiple times, the student model will have better generalization performance than the teacher model. We find that Noisy Student is better with an additional trick: data balancing. Noise Self-training with Noisy Student 1. sign in The abundance of data on the internet is vast. Agreement NNX16AC86A, Is ADS down? The hyperparameters for these noise functions are the same for EfficientNet-B7, L0, L1 and L2. Computer Science - Computer Vision and Pattern Recognition. On ImageNet-P, it leads to an mean flip rate (mFR) of 17.8 if we use a resolution of 224x224 (direct comparison) and 16.1 if we use a resolution of 299x299.111For EfficientNet-L2, we use the model without finetuning with a larger test time resolution, since a larger resolution results in a discrepancy with the resolution of data and leads to degraded performance on ImageNet-C and ImageNet-P. We then perform data filtering and balancing on this corpus. Here we show the evidence in Table 6, noise such as stochastic depth, dropout and data augmentation plays an important role in enabling the student model to perform better than the teacher. Zoph et al. Learn more. Self-training with Noisy Student improves ImageNet classification. Qizhe Xie, Minh-Thang Luong, Eduard Hovy, Quoc V. Le. Code for Noisy Student Training. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. , have shown that computer vision models lack robustness. It extends the idea of self-training and distillation with the use of equal-or-larger student models and noise added to the student during learning. Addressing the lack of robustness has become an important research direction in machine learning and computer vision in recent years. Due to the large model size, the training time of EfficientNet-L2 is approximately five times the training time of EfficientNet-B7. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. For unlabeled images, we set the batch size to be three times the batch size of labeled images for large models, including EfficientNet-B7, L0, L1 and L2. Different types of. However state-of-the-art vision models are still trained with supervised learning which requires a large corpus of labeled images to work well. During the generation of the pseudo labels, the teacher is not noised so that the pseudo labels are as accurate as possible. In our experiments, we also further scale up EfficientNet-B7 and obtain EfficientNet-L0, L1 and L2. Noisy Student Training achieves 88.4% top-1 accuracy on ImageNet, which is 2.0% better than the state-of-the-art model that requires 3.5B weakly labeled Instagram images. unlabeled images , . However, manually annotating organs from CT scans is time . A number of studies, e.g. IEEE Transactions on Pattern Analysis and Machine Intelligence. [76] also proposed to first only train on unlabeled images and then finetune their model on labeled images as the final stage. On robustness test sets, it improves ImageNet-A top-1 accuracy from 61.0% to 83.7%, reduces ImageNet-C mean corruption error from 45.7 to 28.3, and reduces ImageNet-P mean flip rate from 27.8 to 12.2. For labeled images, we use a batch size of 2048 by default and reduce the batch size when we could not fit the model into the memory. With Noisy Student, the model correctly predicts dragonfly for the image. We train our model using the self-training framework[59] which has three main steps: 1) train a teacher model on labeled images, 2) use the teacher to generate pseudo labels on unlabeled images, and 3) train a student model on the combination of labeled images and pseudo labeled images. In the above experiments, iterative training was used to optimize the accuracy of EfficientNet-L2 but here we skip it as it is difficult to use iterative training for many experiments. For instance, on ImageNet-1k, Layer Grafted Pre-training yields 65.5% Top-1 accuracy in terms of 1% few-shot learning with ViT-B/16, which improves MIM and CL baselines by 14.4% and 2.1% with no bells and whistles. on ImageNet, which is 1.0 However an important requirement for Noisy Student to work well is that the student model needs to be sufficiently large to fit more data (labeled and pseudo labeled). Lastly, we apply the recently proposed technique to fix train-test resolution discrepancy[71] for EfficientNet-L0, L1 and L2. For each class, we select at most 130K images that have the highest confidence. [57] used self-training for domain adaptation. This paper presents a unique study of transfer learning with large convolutional networks trained to predict hashtags on billions of social media images and shows improvements on several image classification and object detection tasks, and reports the highest ImageNet-1k single-crop, top-1 accuracy to date. We found that self-training is a simple and effective algorithm to leverage unlabeled data at scale. In all previous experiments, the students capacity is as large as or larger than the capacity of the teacher model. The best model in our experiments is a result of iterative training of teacher and student by putting back the student as the new teacher to generate new pseudo labels. As stated earlier, we hypothesize that noising the student is needed so that it does not merely learn the teachers knowledge. task. Add a By clicking accept or continuing to use the site, you agree to the terms outlined in our. Since a teacher models confidence on an image can be a good indicator of whether it is an out-of-domain image, we consider the high-confidence images as in-domain images and the low-confidence images as out-of-domain images. The Wilds 2.0 update is presented, which extends 8 of the 10 datasets in the Wilds benchmark of distribution shifts to include curated unlabeled data that would be realistically obtainable in deployment, and systematically benchmark state-of-the-art methods that leverage unlabeling data, including domain-invariant, self-training, and self-supervised methods. The main use case of knowledge distillation is model compression by making the student model smaller. After using the masks generated by teacher-SN, the classification performance improved by 0.2 of AC, 1.2 of SP, and 0.7 of AUC. In other words, using Noisy Student makes a much larger impact to the accuracy than changing the architecture. After testing our models robustness to common corruptions and perturbations, we also study its performance on adversarial perturbations. You can also use the colab script noisystudent_svhn.ipynb to try the method on free Colab GPUs. Use Git or checkout with SVN using the web URL. [2] show that Self-Training is superior to Pre-training with ImageNet Supervised Learning on a few Computer . In terms of methodology, We use EfficientNet-B4 as both the teacher and the student. 27.8 to 16.1. On . ImageNet-A test set[25] consists of difficult images that cause significant drops in accuracy to state-of-the-art models. ; 2006)[book reviews], Semi-supervised deep learning with memory, Proceedings of the European Conference on Computer Vision (ECCV), Xception: deep learning with depthwise separable convolutions, K. Clark, M. Luong, C. D. Manning, and Q. V. Le, Semi-supervised sequence modeling with cross-view training, E. D. Cubuk, B. Zoph, D. Mane, V. Vasudevan, and Q. V. Le, AutoAugment: learning augmentation strategies from data, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, E. D. Cubuk, B. Zoph, J. Shlens, and Q. V. Le, RandAugment: practical data augmentation with no separate search, Z. Dai, Z. Yang, F. Yang, W. W. Cohen, and R. R. Salakhutdinov, Good semi-supervised learning that requires a bad gan, T. Furlanello, Z. C. Lipton, M. Tschannen, L. Itti, and A. Anandkumar, A. Galloway, A. Golubeva, T. Tanay, M. Moussa, and G. W. Taylor, R. Geirhos, P. Rubisch, C. Michaelis, M. Bethge, F. A. Wichmann, and W. Brendel, ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness, J. Gilmer, L. Metz, F. Faghri, S. S. Schoenholz, M. Raghu, M. Wattenberg, and I. Goodfellow, I. J. Goodfellow, J. Shlens, and C. Szegedy, Explaining and harnessing adversarial examples, Semi-supervised learning by entropy minimization, Advances in neural information processing systems, K. Gu, B. Yang, J. Ngiam, Q. Especially unlabeled images are plentiful and can be collected with ease. Self-training was previously used to improve ResNet-50 from 76.4% to 81.2% top-1 accuracy[76] which is still far from the state-of-the-art accuracy. https://arxiv.org/abs/1911.04252. The baseline model achieves an accuracy of 83.2. However, the additional hyperparameters introduced by the ramping up schedule and the entropy minimization make them more difficult to use at scale. But during the learning of the student, we inject noise such as data Soft pseudo labels lead to better performance for low confidence data. This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository. Noisy Student leads to significant improvements across all model sizes for EfficientNet. over the JFT dataset to predict a label for each image. We evaluate our EfficientNet-L2 models with and without Noisy Student against an FGSM attack. [^reference-9] [^reference-10] A critical insight was to . As shown in Table3,4 and5, when compared with the previous state-of-the-art model ResNeXt-101 WSL[44, 48] trained on 3.5B weakly labeled images, Noisy Student yields substantial gains on robustness datasets. Le. It is experimentally validated that, for a target test resolution, using a lower train resolution offers better classification at test time, and a simple yet effective and efficient strategy to optimize the classifier performance when the train and test resolutions differ is proposed. Our main results are shown in Table1. Noisy Student Training is a semi-supervised training method which achieves 88.4% top-1 accuracy on ImageNet and surprising gains on robustness and adversarial benchmarks. They did not show significant improvements in terms of robustness on ImageNet-A, C and P as we did. ImageNet-A top-1 accuracy from 16.6 Noisy Student Training is based on the self-training framework and trained with 4 simple steps: Train a classifier on labeled data (teacher). This work proposes a novel architectural unit, which is term the Squeeze-and-Excitation (SE) block, that adaptively recalibrates channel-wise feature responses by explicitly modelling interdependencies between channels and shows that these blocks can be stacked together to form SENet architectures that generalise extremely effectively across different datasets. Use Git or checkout with SVN using the web URL. Are you sure you want to create this branch? Our study shows that using unlabeled data improves accuracy and general robustness. Scripts used for our ImageNet experiments: Similar scripts to run predictions on unlabeled data, filter and balance data and train using the filtered data. The ADS is operated by the Smithsonian Astrophysical Observatory under NASA Cooperative On ImageNet, we first train an EfficientNet model on labeled images and use it as a teacher to generate pseudo labels for 300M unlabeled images. Our finding is consistent with similar arguments that using unlabeled data can improve adversarial robustness[8, 64, 46, 80]. . In our experiments, we use dropout[63], stochastic depth[29], data augmentation[14] to noise the student. As noise injection methods are not used in the student model, and the student model was also small, it is more difficult to make the student better than teacher. The architectures for the student and teacher models can be the same or different. Please Notably, EfficientNet-B7 achieves an accuracy of 86.8%, which is 1.8% better than the supervised model. We will then show our results on ImageNet and compare them with state-of-the-art models. To noise the student, we use dropout[63], data augmentation[14] and stochastic depth[29] during its training. Self-Training With Noisy Student Improves ImageNet Classification @article{Xie2019SelfTrainingWN, title={Self-Training With Noisy Student Improves ImageNet Classification}, author={Qizhe Xie and Eduard H. Hovy and Minh-Thang Luong and Quoc V. Le}, journal={2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, year={2019 . On robustness test sets, it improves ImageNet-A top-1 accuracy from 61.0% to 83.7%, reduces ImageNet-C mean corruption error from 45.7 to 28.3, and reduces ImageNet-P mean flip rate from 27.8 to 12.2. team using this approach not only surpasses the top-1 ImageNet accuracy of SOTA models by 1%, it also shows that the robustness of a model also improves. To achieve this result, we first train an EfficientNet model on labeled ImageNet images and use it as a teacher to generate pseudo labels on 300M unlabeled images. Noisy Student Training achieves 88.4% top-1 accuracy on ImageNet, which is 2.0% better than the state-of-the-art model that requires 3.5B weakly labeled Instagram images. Noisy Student (B7, L2) means to use EfficientNet-B7 as the student and use our best model with 87.4% accuracy as the teacher model. Self-training 1 2Self-training 3 4n What is Noisy Student? Since we use soft pseudo labels generated from the teacher model, when the student is trained to be exactly the same as the teacher model, the cross entropy loss on unlabeled data would be zero and the training signal would vanish. Abdominal organ segmentation is very important for clinical applications. We duplicate images in classes where there are not enough images. ImageNet images and use it as a teacher to generate pseudo labels on 300M On ImageNet, we first train an EfficientNet model on labeled images and use it as a teacher to generate pseudo labels for 300M unlabeled images. Iterative training is not used here for simplicity. We use our best model Noisy Student with EfficientNet-L2 to teach student models with sizes ranging from EfficientNet-B0 to EfficientNet-B7. et al. Parthasarathi et al. Self-training with Noisy Student improves ImageNet classification. IEEE Trans. Figure 1(b) shows images from ImageNet-C and the corresponding predictions.