The document discusses an experiment comparing the transfer learning performance of standard ImageNet models versus adversarially robust ImageNet models. The experiment finds that robust models consistently match or outperform standard models on a variety of downstream transfer learning tasks, despite having lower accuracy on ImageNet. Further analysis shows robust models improve with increased width and that the optimal level of robustness depends on properties of the downstream task like dataset granularity. Overall, the findings suggest adversarially robust models transfer learned representations better than standard models.
PR-330: How To Train Your ViT? Data, Augmentation, and Regularization in Visi...Jinwon Lee
안녕하세요 TensorFlow Korea 논문 읽기 모임 PR-12의 330번째 논문 리뷰입니다.
오늘은 무려 5만개의 학습된 ViT model을 제공하는 구글스러운 논문을 리뷰해보았습니다. ViT가 CNN을 조금씩 대체해가고 있는데요, ViT는 CNN과 달리 inductive bias가 적은 관계로
좋은 성능을 위해서는 굉장히 많은 data가 필요하거나, augmentation과 regularization을 많이 써줘야 합니다.
그런데 이렇게 다양한 경우 즉 다양한 data, 다양한 model size, 다양한 augmentation 방법, 다양한 regularization, 다양한 data size 등등에 따른 ViT의 성능과 속도 등의 비교 분석 실험이 지금까지는 없었죠.
이 논문에서는 그 어려운 걸(?) 해냈습니다. 그리고 수많은 ViT를 이용해 실험을 하면서 몇가지 중요한 finding들을 찾았습니다.
요약하면 다음과 같습니다.
1. augmentation과 regularization을 잘 쓰면 1/10의 data로도 전체 data 다 쓴거랑 대부분 비슷한 성능을 낼 수 있다. 그런데 항상 그런건 아니다.
반대로 말하면 data가 10배 있으면 augmentation이나 regularization안 쓰고도 좋은 성능을 낼 수 있다.
2. downstream task 학습할 때 scratch부터 학습하는거랑 large dataset으로 pre-trained한 걸 이용해서 transfer learning하는 건 후자가 좋다.
3. transfer learning 할 때도 pre-trained model 중에 data 많이 써서 학습한게 더 좋다.
4. augmentation/regularization은 data가 많으면 별 도움이 안되고 둘 중에는 augmenation이 더 좋다.
5. pre-trained model이 많을 때 model을 고르는 방법은 그냥 upstream에서 제일 잘됐던 걸 고르면 얼추 잘된다.
6. 속도를 빠르게 하고 싶을 때는 model을 작은거 쓰지말고 patch size를 키워라. 그래야 성능이 별로 안떨어진다.
입니다.
흥미로운 결과들이 많으니 자세한 내용은 아래 영상을 참고해주세요!
감사합니다!
영상링크: https://youtu.be/A3RrAIx-KCc
논문링크: https://arxiv.org/abs/2106.10270
PR-231: A Simple Framework for Contrastive Learning of Visual RepresentationsJinwon Lee
TensorFlow Korea 논문읽기모임 PR12 231번째 논문 review 입니다
이번 논문은 Google Brain에서 나온 A Simple Framework for Contrastive Learning of Visual Representations입니다. Geoffrey Hinton님이 마지막 저자이시기도 해서 최근에 더 주목을 받고 있는 논문입니다.
이 논문은 최근에 굉장히 핫한 topic인 contrastive learning을 이용한 self-supervised learning쪽 논문으로 supervised learning으로 학습한 ResNet50와 동일한 성능을 얻을 수 있는 unsupervised pre-trainig 방법을 제안하였습니다. Data augmentation, Non-linear projection head, large batch size, longer training, NTXent loss 등을 활용하여 훌륭한 representation learning이 가능함을 보여주었고, semi-supervised learning이나 transfer learning에서도 매우 뛰어난 결과를 보여주었습니다. 자세한 내용은 영상을 참고해주세요
논문링크: https://arxiv.org/abs/2002.05709
영상링크: https://youtu.be/FWhM3juUM6s
PR-297: Training data-efficient image transformers & distillation through att...Jinwon Lee
안녕하세요 TensorFlow Korea 논문 읽기 모임 PR-12의 297번째 리뷰입니다
어느덧 PR-12 시즌 3의 끝까지 논문 3편밖에 남지 않았네요.
시즌 3가 끝나면 바로 시즌 4의 새 멤버 모집이 시작될 예정입니다. 많은 관심과 지원 부탁드립니다~~
(멤버 모집 공지는 Facebook TensorFlow Korea 그룹에 올라올 예정입니다)
오늘 제가 리뷰한 논문은 Facebook의 Training data-efficient image transformers & distillation through attention 입니다.
Google에서 나왔던 ViT논문 이후에 convolution을 전혀 사용하지 않고 오직 attention만을 이용한 computer vision algorithm에 어느때보다 관심이 높아지고 있는데요
이 논문에서 제안한 DeiT 모델은 ViT와 같은 architecture를 사용하면서 ViT가 ImageNet data만으로는 성능이 잘 안나왔던 것에 비해서
Training 방법 개선과 새로운 Knowledge Distillation 방법을 사용하여 mageNet data 만으로 EfficientNet보다 뛰어난 성능을 보여주는 결과를 얻었습니다.
정말 CNN은 이제 서서히 사라지게 되는 것일까요? Attention이 computer vision도 정복하게 될 것인지....
개인적으로는 당분간은 attention 기반의 CV 논문이 쏟아질 거라고 확신하고, 또 여기에서 놀라운 일들이 일어날 수 있을 거라고 생각하고 있습니다
CNN은 10년간 많은 연구를 통해서 발전해왔지만, transformer는 이제 CV에 적용된 지 얼마 안된 시점이라서 더 기대가 크구요,
attention이 inductive bias가 가장 적은 형태의 모델이기 때문에 더 놀라운 이들을 만들 수 있을거라고 생각합니다
얼마 전에 나온 open AI의 DALL-E도 그 대표적인 예라고 할 수 있을 것 같습니다. Transformer의 또하나의 transformation이 궁금하신 분들은 아래 영상을 참고해주세요
영상링크: https://youtu.be/DjEvzeiWBTo
논문링크: https://arxiv.org/abs/2012.12877
For the full video of this presentation, please visit: https://www.edge-ai-vision.com/2021/10/dnn-training-data-how-to-know-what-you-need-and-how-to-get-it-a-presentation-from-tech-mahindra/
Abhishek Sharma, Practice Head for Engineering AI at Tech Mahindra, presents the “DNN Training Data: How to Know What You Need and How to Get It” tutorial at the May 2021 Embedded Vision Summit.
Successful training of deep neural networks requires the right amounts and types of annotated training data. Collecting, curating and labeling this data is typically one of the most time-consuming aspects of developing a deep-learning-based solution.
In this talk, Sharma discusses approaches useful for situations where insufficient data is available, including transfer learning and data augmentation, including the use of generative adversarial networks (GANs). He also discusses techniques that can be helpful when data is plentiful, such as transforms, data path optimization and approximate computing. He illustrates these techniques and challenges via case studies from the healthcare and manufacturing industries.
PR-270: PP-YOLO: An Effective and Efficient Implementation of Object DetectorJinwon Lee
TensorFlow Korea 논문읽기모임 PR12 270번째 논문 review입니다.
이번 논문은 Baidu에서 나온 PP-YOLO: An Effective and Efficient Implementation of Object Detector입니다. YOLOv3에 다양한 방법을 적용하여 매우 높은 성능과 함께 매우 빠른 속도 두마리 토끼를 다 잡아버린(?) 그런 논문입니다. 논문에서 사용한 다양한 trick들에 대해서 좀 더 깊이있게 살펴보았습니다. Object detection에 사용된 기법 들 중에 Deformable convolution, Exponential Moving Average, DropBlock, IoU aware prediction, Grid sensitivity elimination, MatrixNMS, CoordConv, 등의 방법에 관심이 있으시거나 알고 싶으신 분들은 영상과 발표자료를 참고하시면 좋을 것 같습니다!
논문링크: https://arxiv.org/abs/2007.12099
영상링크: https://youtu.be/7v34cCE5H4k
For real world application, convolutional neural network(CNN) model can take more than 100MB of space and can be computationally too expensive. Therefore, there are multiple methods to reduce this complexity in the state of art. Ristretto is a plug-in to Caffe framework that employs several model approximation methods. For this projects, first a CNN model is trained for Cifar-10 dataset with Caffe, then Ristretto will be use to generate multiple approximated version of the trained model using different schemes. The goal of this projects is comparison of the models in terms of execution performance, model size and cache utilizations in the test or inference phase. The same steps are done with Tensorflow and Quantisation tool. The quantisation schemes of Tensorflow and Ristretto are then compared.
PR-217: EfficientDet: Scalable and Efficient Object DetectionJinwon Lee
TensorFlow Korea 논문읽기모임 PR12 217번째 논문 review입니다
이번 논문은 GoogleBrain에서 쓴 EfficientDet입니다. EfficientNet의 후속작으로 accuracy와 efficiency를 둘 다 잡기 위한 object detection 방법을 제안한 논문입니다. 이를 위하여 weighted bidirectional feature pyramid network(BiFPN)과 EfficientNet과 유사한 방법의 detection용 compound scaling 방법을 제안하고 있는데요, 자세한 내용은 영상을 참고해주세요
논문링크: https://arxiv.org/abs/1911.09070
영상링크: https://youtu.be/11jDC8uZL0E
[PR12] Inception and Xception - Jaejun YooJaeJun Yoo
Introduction to Inception and Xception
video: https://youtu.be/V0dLhyg5_Dw
Papers:
Going Deeper with Convolutions
Rethinking the Inception Architecture for Computer Vision
Inception-v4, Inception-RestNet and the Impact of Residual Connections on Learning
Xception: Deep Learning with Depthwise Separable Convolutions
For the full video of this presentation, please visit:
https://www.embedded-vision.com/platinum-members/embedded-vision-alliance/embedded-vision-training/videos/pages/sep-2019-alliance-vitf-facebook
For more information about embedded vision, please visit:
http://www.embedded-vision.com
Raghuraman Krishnamoorthi, Software Engineer at Facebook, delivers the presentation "Quantizing Deep Networks for Efficient Inference at the Edge" at the Embedded Vision Alliance's September 2019 Vision Industry and Technology Forum. Krishnamoorthi gives an overview of practical deep neural network quantization techniques and tools.
Task Adaptive Neural Network Search with Meta-Contrastive LearningMLAI2
Most conventional Neural Architecture Search (NAS) approaches are limited in that they only generate architectures without searching for the optimal parameters. While some NAS methods handle this issue by utilizing a supernet trained on a large-scale dataset such as ImageNet, they may be suboptimal if the target tasks are highly dissimilar from the dataset the supernet is trained on. To address such limitations, we introduce a novel problem of Neural Network Search (NNS), whose goal is to search for the optimal pretrained network for a novel dataset and constraints (e.g. number of parameters), from a model zoo. Then, we propose a novel framework to tackle the problem, namely Task-Adaptive Neural Network Search (TANS). Given a model-zoo that consists of network pretrained on diverse datasets, we use a novel amortized meta-learning framework to learn a cross-modal latent space with contrastive loss, to maximize the similarity between a dataset and a high-performing network on it, and minimize the similarity between irrelevant dataset-network pairs. We validate the effectiveness and efficiency of our method on ten real-world datasets, against existing NAS/AutoML baselines. The results show that our method instantly retrieves networks that outperform models obtained with the baselines with significantly fewer training steps to reach the target performance, thus minimizing the total cost of obtaining a task-optimal network. Our code and the model-zoo are available at https://anonymous.4open.science/r/TANS-33D6
PR-330: How To Train Your ViT? Data, Augmentation, and Regularization in Visi...Jinwon Lee
안녕하세요 TensorFlow Korea 논문 읽기 모임 PR-12의 330번째 논문 리뷰입니다.
오늘은 무려 5만개의 학습된 ViT model을 제공하는 구글스러운 논문을 리뷰해보았습니다. ViT가 CNN을 조금씩 대체해가고 있는데요, ViT는 CNN과 달리 inductive bias가 적은 관계로
좋은 성능을 위해서는 굉장히 많은 data가 필요하거나, augmentation과 regularization을 많이 써줘야 합니다.
그런데 이렇게 다양한 경우 즉 다양한 data, 다양한 model size, 다양한 augmentation 방법, 다양한 regularization, 다양한 data size 등등에 따른 ViT의 성능과 속도 등의 비교 분석 실험이 지금까지는 없었죠.
이 논문에서는 그 어려운 걸(?) 해냈습니다. 그리고 수많은 ViT를 이용해 실험을 하면서 몇가지 중요한 finding들을 찾았습니다.
요약하면 다음과 같습니다.
1. augmentation과 regularization을 잘 쓰면 1/10의 data로도 전체 data 다 쓴거랑 대부분 비슷한 성능을 낼 수 있다. 그런데 항상 그런건 아니다.
반대로 말하면 data가 10배 있으면 augmentation이나 regularization안 쓰고도 좋은 성능을 낼 수 있다.
2. downstream task 학습할 때 scratch부터 학습하는거랑 large dataset으로 pre-trained한 걸 이용해서 transfer learning하는 건 후자가 좋다.
3. transfer learning 할 때도 pre-trained model 중에 data 많이 써서 학습한게 더 좋다.
4. augmentation/regularization은 data가 많으면 별 도움이 안되고 둘 중에는 augmenation이 더 좋다.
5. pre-trained model이 많을 때 model을 고르는 방법은 그냥 upstream에서 제일 잘됐던 걸 고르면 얼추 잘된다.
6. 속도를 빠르게 하고 싶을 때는 model을 작은거 쓰지말고 patch size를 키워라. 그래야 성능이 별로 안떨어진다.
입니다.
흥미로운 결과들이 많으니 자세한 내용은 아래 영상을 참고해주세요!
감사합니다!
영상링크: https://youtu.be/A3RrAIx-KCc
논문링크: https://arxiv.org/abs/2106.10270
PR-231: A Simple Framework for Contrastive Learning of Visual RepresentationsJinwon Lee
TensorFlow Korea 논문읽기모임 PR12 231번째 논문 review 입니다
이번 논문은 Google Brain에서 나온 A Simple Framework for Contrastive Learning of Visual Representations입니다. Geoffrey Hinton님이 마지막 저자이시기도 해서 최근에 더 주목을 받고 있는 논문입니다.
이 논문은 최근에 굉장히 핫한 topic인 contrastive learning을 이용한 self-supervised learning쪽 논문으로 supervised learning으로 학습한 ResNet50와 동일한 성능을 얻을 수 있는 unsupervised pre-trainig 방법을 제안하였습니다. Data augmentation, Non-linear projection head, large batch size, longer training, NTXent loss 등을 활용하여 훌륭한 representation learning이 가능함을 보여주었고, semi-supervised learning이나 transfer learning에서도 매우 뛰어난 결과를 보여주었습니다. 자세한 내용은 영상을 참고해주세요
논문링크: https://arxiv.org/abs/2002.05709
영상링크: https://youtu.be/FWhM3juUM6s
PR-297: Training data-efficient image transformers & distillation through att...Jinwon Lee
안녕하세요 TensorFlow Korea 논문 읽기 모임 PR-12의 297번째 리뷰입니다
어느덧 PR-12 시즌 3의 끝까지 논문 3편밖에 남지 않았네요.
시즌 3가 끝나면 바로 시즌 4의 새 멤버 모집이 시작될 예정입니다. 많은 관심과 지원 부탁드립니다~~
(멤버 모집 공지는 Facebook TensorFlow Korea 그룹에 올라올 예정입니다)
오늘 제가 리뷰한 논문은 Facebook의 Training data-efficient image transformers & distillation through attention 입니다.
Google에서 나왔던 ViT논문 이후에 convolution을 전혀 사용하지 않고 오직 attention만을 이용한 computer vision algorithm에 어느때보다 관심이 높아지고 있는데요
이 논문에서 제안한 DeiT 모델은 ViT와 같은 architecture를 사용하면서 ViT가 ImageNet data만으로는 성능이 잘 안나왔던 것에 비해서
Training 방법 개선과 새로운 Knowledge Distillation 방법을 사용하여 mageNet data 만으로 EfficientNet보다 뛰어난 성능을 보여주는 결과를 얻었습니다.
정말 CNN은 이제 서서히 사라지게 되는 것일까요? Attention이 computer vision도 정복하게 될 것인지....
개인적으로는 당분간은 attention 기반의 CV 논문이 쏟아질 거라고 확신하고, 또 여기에서 놀라운 일들이 일어날 수 있을 거라고 생각하고 있습니다
CNN은 10년간 많은 연구를 통해서 발전해왔지만, transformer는 이제 CV에 적용된 지 얼마 안된 시점이라서 더 기대가 크구요,
attention이 inductive bias가 가장 적은 형태의 모델이기 때문에 더 놀라운 이들을 만들 수 있을거라고 생각합니다
얼마 전에 나온 open AI의 DALL-E도 그 대표적인 예라고 할 수 있을 것 같습니다. Transformer의 또하나의 transformation이 궁금하신 분들은 아래 영상을 참고해주세요
영상링크: https://youtu.be/DjEvzeiWBTo
논문링크: https://arxiv.org/abs/2012.12877
For the full video of this presentation, please visit: https://www.edge-ai-vision.com/2021/10/dnn-training-data-how-to-know-what-you-need-and-how-to-get-it-a-presentation-from-tech-mahindra/
Abhishek Sharma, Practice Head for Engineering AI at Tech Mahindra, presents the “DNN Training Data: How to Know What You Need and How to Get It” tutorial at the May 2021 Embedded Vision Summit.
Successful training of deep neural networks requires the right amounts and types of annotated training data. Collecting, curating and labeling this data is typically one of the most time-consuming aspects of developing a deep-learning-based solution.
In this talk, Sharma discusses approaches useful for situations where insufficient data is available, including transfer learning and data augmentation, including the use of generative adversarial networks (GANs). He also discusses techniques that can be helpful when data is plentiful, such as transforms, data path optimization and approximate computing. He illustrates these techniques and challenges via case studies from the healthcare and manufacturing industries.
PR-270: PP-YOLO: An Effective and Efficient Implementation of Object DetectorJinwon Lee
TensorFlow Korea 논문읽기모임 PR12 270번째 논문 review입니다.
이번 논문은 Baidu에서 나온 PP-YOLO: An Effective and Efficient Implementation of Object Detector입니다. YOLOv3에 다양한 방법을 적용하여 매우 높은 성능과 함께 매우 빠른 속도 두마리 토끼를 다 잡아버린(?) 그런 논문입니다. 논문에서 사용한 다양한 trick들에 대해서 좀 더 깊이있게 살펴보았습니다. Object detection에 사용된 기법 들 중에 Deformable convolution, Exponential Moving Average, DropBlock, IoU aware prediction, Grid sensitivity elimination, MatrixNMS, CoordConv, 등의 방법에 관심이 있으시거나 알고 싶으신 분들은 영상과 발표자료를 참고하시면 좋을 것 같습니다!
논문링크: https://arxiv.org/abs/2007.12099
영상링크: https://youtu.be/7v34cCE5H4k
For real world application, convolutional neural network(CNN) model can take more than 100MB of space and can be computationally too expensive. Therefore, there are multiple methods to reduce this complexity in the state of art. Ristretto is a plug-in to Caffe framework that employs several model approximation methods. For this projects, first a CNN model is trained for Cifar-10 dataset with Caffe, then Ristretto will be use to generate multiple approximated version of the trained model using different schemes. The goal of this projects is comparison of the models in terms of execution performance, model size and cache utilizations in the test or inference phase. The same steps are done with Tensorflow and Quantisation tool. The quantisation schemes of Tensorflow and Ristretto are then compared.
PR-217: EfficientDet: Scalable and Efficient Object DetectionJinwon Lee
TensorFlow Korea 논문읽기모임 PR12 217번째 논문 review입니다
이번 논문은 GoogleBrain에서 쓴 EfficientDet입니다. EfficientNet의 후속작으로 accuracy와 efficiency를 둘 다 잡기 위한 object detection 방법을 제안한 논문입니다. 이를 위하여 weighted bidirectional feature pyramid network(BiFPN)과 EfficientNet과 유사한 방법의 detection용 compound scaling 방법을 제안하고 있는데요, 자세한 내용은 영상을 참고해주세요
논문링크: https://arxiv.org/abs/1911.09070
영상링크: https://youtu.be/11jDC8uZL0E
[PR12] Inception and Xception - Jaejun YooJaeJun Yoo
Introduction to Inception and Xception
video: https://youtu.be/V0dLhyg5_Dw
Papers:
Going Deeper with Convolutions
Rethinking the Inception Architecture for Computer Vision
Inception-v4, Inception-RestNet and the Impact of Residual Connections on Learning
Xception: Deep Learning with Depthwise Separable Convolutions
For the full video of this presentation, please visit:
https://www.embedded-vision.com/platinum-members/embedded-vision-alliance/embedded-vision-training/videos/pages/sep-2019-alliance-vitf-facebook
For more information about embedded vision, please visit:
http://www.embedded-vision.com
Raghuraman Krishnamoorthi, Software Engineer at Facebook, delivers the presentation "Quantizing Deep Networks for Efficient Inference at the Edge" at the Embedded Vision Alliance's September 2019 Vision Industry and Technology Forum. Krishnamoorthi gives an overview of practical deep neural network quantization techniques and tools.
Task Adaptive Neural Network Search with Meta-Contrastive LearningMLAI2
Most conventional Neural Architecture Search (NAS) approaches are limited in that they only generate architectures without searching for the optimal parameters. While some NAS methods handle this issue by utilizing a supernet trained on a large-scale dataset such as ImageNet, they may be suboptimal if the target tasks are highly dissimilar from the dataset the supernet is trained on. To address such limitations, we introduce a novel problem of Neural Network Search (NNS), whose goal is to search for the optimal pretrained network for a novel dataset and constraints (e.g. number of parameters), from a model zoo. Then, we propose a novel framework to tackle the problem, namely Task-Adaptive Neural Network Search (TANS). Given a model-zoo that consists of network pretrained on diverse datasets, we use a novel amortized meta-learning framework to learn a cross-modal latent space with contrastive loss, to maximize the similarity between a dataset and a high-performing network on it, and minimize the similarity between irrelevant dataset-network pairs. We validate the effectiveness and efficiency of our method on ten real-world datasets, against existing NAS/AutoML baselines. The results show that our method instantly retrieves networks that outperform models obtained with the baselines with significantly fewer training steps to reach the target performance, thus minimizing the total cost of obtaining a task-optimal network. Our code and the model-zoo are available at https://anonymous.4open.science/r/TANS-33D6
How useful is self-supervised pretraining for Visual tasks?Seunghyun Hwang
Review : How useful is self-supervised pretraining for Visual tasks?
- by Seunghyun Hwang (Yonsei University, Severance Hospital, Center for Clinical Data Science)
Multi-modal sources for predictive modeling using deep learningSanghamitra Deb
Using Vision Language models : Is it possible to prompt them similar to LLMs? when to use out of the box and when to pre-train? General multi-modal models --- deeplearning. Machine learning metrics, feature engineering and setting up an ML problem.
Unsupervised generative methods have undergone a recent renaissance, spurred on in large part by impressive photo-realistic results in image applications. These generative methods seek to yield models that understand data by learning how to generate samples through implicit and explicit likelihood optimization. However, despite the surge in interest, these models are limited in several key aspects. First, although methods with an explicit likelihood are, in principle, able to perform additional tasks like anomaly detection and imputation, biases in the learned likelihood render these models useless for such important tasks. For example, recent work has shown that modern methods lead to high out-of-distribution likelihoods for data that is unlike seen training instances. Secondly, most current generative methods are limited to fixed-length vector or sequential data, leaving a substantial gap for the analysis of exchangeable data like sets and graphs. I.e., modern generative models excel at modeling dependencies among features in a point, but are lacking in modeling dependencies amongpoints in a collection. In this talk I discuss these shortcomings and suggest some possible avenues for improvement.
To Get any Project for CSE, IT ECE, EEE Contact Me @ 09666155510, 09849539085 or mail us - ieeefinalsemprojects@gmail.com-Visit Our Website: www.finalyearprojects.org
To Get any Project for CSE, IT ECE, EEE Contact Me @ 09666155510, 09849539085 or mail us - ieeefinalsemprojects@gmail.com-Visit Our Website: www.finalyearprojects.org
To Get any Project for CSE, IT ECE, EEE Contact Me @ 09666155510, 09849539085 or mail us - ieeefinalsemprojects@gmail.com-Visit Our Website: www.finalyearprojects.org
PR-169: EfficientNet: Rethinking Model Scaling for Convolutional Neural NetworksJinwon Lee
TensorFlow-KR 논문읽기모임 PR12 169번째 논문 review입니다.
이번에 살펴본 논문은 Google에서 발표한 EfficientNet입니다. efficient neural network은 보통 mobile과 같은 제한된 computing power를 가진 edge device를 위한 작은 network 위주로 연구되어왔는데, 이 논문은 성능을 높이기 위해서 일반적으로 network를 점점 더 키워나가는 경우가 많은데, 이 때 어떻게 하면 더 효율적인 방법으로 network을 키울 수 있을지에 대해서 연구한 논문입니다. 자세한 내용은 영상을 참고해주세요
논문링크: https://arxiv.org/abs/1905.11946
영상링크: https://youtu.be/Vhz0quyvR7I
We presents a deep architecture for dense semantic correspondence, called pyramidal affine regression networks (PARN), that estimates locally-varying affine transformation fields across images.
To deal with intra-class appearance and shape variations that commonly exist among different instances within the same object category,
we leverage a pyramidal model where affine transformation fields are progressively estimated in a coarse-to-fine manner so that the smoothness constraint is naturally imposed within deep networks.
PARN estimates residual affine transformations at each level and composes them to estimate final affine transformations.
Furthermore, to overcome the limitations of insufficient training data for semantic correspondence, we propose a novel weakly-supervised training scheme that generates progressive supervisions by leveraging a correspondence consistency across image pairs.
Our method is fully learnable in an end-to-end manner and does not require quantizing infinite continuous affine transformation fields.
Computer Vision abbreviated as CV aims to teach computers to achieve human level vision capabilities. Applications of CV in self driving cars, robotics, healthcare, education and the multitude of apps that allow customers to use the smartphone cameras to convey information has made it one of the most popular fields in Artificial Intelligence. The recent advances in Deep Learning, data storage and computing capabilities has lead to the huge success of CV. There are several tasks in computer vision, such as classification, object detection, image segmentation, optical character recognition, scene reconstruction and many others.
In this presentation I will talk about applying Transfer Learning, Image classification, object detection and the metrics required to measure them on still images. The increase in accuracy over of CV tasks over the past decade is due to Convolutional Neural Networks (CNN), CNN is the base used in architectures such as RESNET or VGGNET. I will go through how to use these pre-trained models for image classification and feature extraction. One of the break throughs in object detection has come with one-shot learning, where the bounding box and the class of the object is predicted simultaneously. This leads to low latency during inference (155 frames per second) and high accuracy. This is the framework behind object detection using YOLO , I will explain how to use yolo for specific use cases.
In this video from the ISC Big Data'14 Conference, Ted Willke from Intel presents: The Analytics Frontier of the Hadoop Eco-System.
"The Hadoop MapReduce framework grew out of an effort to make it easy to express and parallelize simple computations that were routinely performed at Google. It wasn’t long before libraries, like Apache Mahout, were developed to enable matrix factorization, clustering, regression, and other more complex analyses on Hadoop. Now, many of these libraries and their workloads are migrating to Apache Spark because it supports a wider class of applications than MapReduce and is more appropriate for iterative algorithms, interactive processing, and streaming applications. What’s next beyond Spark? Where is big data analytics processing headed? How will data scientists program these systems? In this talk, we will explore the current analytics frontier, the popular debates, and discuss some potentially clever additions. We will also share the emergent data science applications and collaborative university research that inform our thinking."
Learn more:
http://www.isc-events.com/bigdata14/schedule.html
and
http://www.intel.com/content/www/us/en/software/intel-graph-solutions.html
Watch the video presentation: https://www.youtube.com/watch?v=qlfx495Ekw0
GraphRAG is All You need? LLM & Knowledge GraphGuy Korland
Guy Korland, CEO and Co-founder of FalkorDB, will review two articles on the integration of language models with knowledge graphs.
1. Unifying Large Language Models and Knowledge Graphs: A Roadmap.
https://arxiv.org/abs/2306.08302
2. Microsoft Research's GraphRAG paper and a review paper on various uses of knowledge graphs:
https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/
Smart TV Buyer Insights Survey 2024 by 91mobiles.pdf91mobiles
91mobiles recently conducted a Smart TV Buyer Insights Survey in which we asked over 3,000 respondents about the TV they own, aspects they look at on a new TV, and their TV buying preferences.
Elevating Tactical DDD Patterns Through Object CalisthenicsDorra BARTAGUIZ
After immersing yourself in the blue book and its red counterpart, attending DDD-focused conferences, and applying tactical patterns, you're left with a crucial question: How do I ensure my design is effective? Tactical patterns within Domain-Driven Design (DDD) serve as guiding principles for creating clear and manageable domain models. However, achieving success with these patterns requires additional guidance. Interestingly, we've observed that a set of constraints initially designed for training purposes remarkably aligns with effective pattern implementation, offering a more ‘mechanical’ approach. Let's explore together how Object Calisthenics can elevate the design of your tactical DDD patterns, offering concrete help for those venturing into DDD for the first time!
Software Delivery At the Speed of AI: Inflectra Invests In AI-Powered QualityInflectra
In this insightful webinar, Inflectra explores how artificial intelligence (AI) is transforming software development and testing. Discover how AI-powered tools are revolutionizing every stage of the software development lifecycle (SDLC), from design and prototyping to testing, deployment, and monitoring.
Learn about:
• The Future of Testing: How AI is shifting testing towards verification, analysis, and higher-level skills, while reducing repetitive tasks.
• Test Automation: How AI-powered test case generation, optimization, and self-healing tests are making testing more efficient and effective.
• Visual Testing: Explore the emerging capabilities of AI in visual testing and how it's set to revolutionize UI verification.
• Inflectra's AI Solutions: See demonstrations of Inflectra's cutting-edge AI tools like the ChatGPT plugin and Azure Open AI platform, designed to streamline your testing process.
Whether you're a developer, tester, or QA professional, this webinar will give you valuable insights into how AI is shaping the future of software delivery.
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...Ramesh Iyer
In today's fast-changing business world, Companies that adapt and embrace new ideas often need help to keep up with the competition. However, fostering a culture of innovation takes much work. It takes vision, leadership and willingness to take risks in the right proportion. Sachin Dev Duggal, co-founder of Builder.ai, has perfected the art of this balance, creating a company culture where creativity and growth are nurtured at each stage.
GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...James Anderson
Effective Application Security in Software Delivery lifecycle using Deployment Firewall and DBOM
The modern software delivery process (or the CI/CD process) includes many tools, distributed teams, open-source code, and cloud platforms. Constant focus on speed to release software to market, along with the traditional slow and manual security checks has caused gaps in continuous security as an important piece in the software supply chain. Today organizations feel more susceptible to external and internal cyber threats due to the vast attack surface in their applications supply chain and the lack of end-to-end governance and risk management.
The software team must secure its software delivery process to avoid vulnerability and security breaches. This needs to be achieved with existing tool chains and without extensive rework of the delivery processes. This talk will present strategies and techniques for providing visibility into the true risk of the existing vulnerabilities, preventing the introduction of security issues in the software, resolving vulnerabilities in production environments quickly, and capturing the deployment bill of materials (DBOM).
Speakers:
Bob Boule
Robert Boule is a technology enthusiast with PASSION for technology and making things work along with a knack for helping others understand how things work. He comes with around 20 years of solution engineering experience in application security, software continuous delivery, and SaaS platforms. He is known for his dynamic presentations in CI/CD and application security integrated in software delivery lifecycle.
Gopinath Rebala
Gopinath Rebala is the CTO of OpsMx, where he has overall responsibility for the machine learning and data processing architectures for Secure Software Delivery. Gopi also has a strong connection with our customers, leading design and architecture for strategic implementations. Gopi is a frequent speaker and well-known leader in continuous delivery and integrating security into software delivery.
DevOps and Testing slides at DASA ConnectKari Kakkonen
My and Rik Marselis slides at 30.5.2024 DASA Connect conference. We discuss about what is testing, then what is agile testing and finally what is Testing in DevOps. Finally we had lovely workshop with the participants trying to find out different ways to think about quality and testing in different parts of the DevOps infinity loop.
Transcript: Selling digital books in 2024: Insights from industry leaders - T...BookNet Canada
The publishing industry has been selling digital audiobooks and ebooks for over a decade and has found its groove. What’s changed? What has stayed the same? Where do we go from here? Join a group of leading sales peers from across the industry for a conversation about the lessons learned since the popularization of digital books, best practices, digital book supply chain management, and more.
Link to video recording: https://bnctechforum.ca/sessions/selling-digital-books-in-2024-insights-from-industry-leaders/
Presented by BookNet Canada on May 28, 2024, with support from the Department of Canadian Heritage.
Connector Corner: Automate dynamic content and events by pushing a buttonDianaGray10
Here is something new! In our next Connector Corner webinar, we will demonstrate how you can use a single workflow to:
Create a campaign using Mailchimp with merge tags/fields
Send an interactive Slack channel message (using buttons)
Have the message received by managers and peers along with a test email for review
But there’s more:
In a second workflow supporting the same use case, you’ll see:
Your campaign sent to target colleagues for approval
If the “Approve” button is clicked, a Jira/Zendesk ticket is created for the marketing design team
But—if the “Reject” button is pushed, colleagues will be alerted via Slack message
Join us to learn more about this new, human-in-the-loop capability, brought to you by Integration Service connectors.
And...
Speakers:
Akshay Agnihotri, Product Manager
Charlie Greenberg, Host
Accelerate your Kubernetes clusters with Varnish CachingThijs Feryn
A presentation about the usage and availability of Varnish on Kubernetes. This talk explores the capabilities of Varnish caching and shows how to use the Varnish Helm chart to deploy it to Kubernetes.
This presentation was delivered at K8SUG Singapore. See https://feryn.eu/presentations/accelerate-your-kubernetes-clusters-with-varnish-caching-k8sug-singapore-28-2024 for more details.
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024Albert Hoitingh
In this session I delve into the encryption technology used in Microsoft 365 and Microsoft Purview. Including the concepts of Customer Key and Double Key Encryption.
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
do adversarially robust image net models transfer better
1. 2020/11/29
Ho Seong Lee (hoya012)
Cognex Deep Learning Lab
Research Engineer
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 1
2. Contents
• Introduction
• Related Work
• Experiments
• Analysis & Discussion
• Conclusion
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 2
3. Introduction
Transfer Learning is a widely-used paradigm in deep learning (maybe.. default..?)
• Models pre-trained on standard datasets (e.g. ImageNet) can be efficiently adapted to downstream tasks.
• Better pre-trained models yield better transfer results, suggesting that initial accuracy is a key aspect of
transfer learning performance.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 3
Reference: “Do Better ImageNet Models Transfer Better?“, 2019 CVPR
4. Related Works
Transfer Learning in various domain
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 4
• Medical imaging
• “Comparison of deep transfer learning strategies for digital pathology”, 2018 CVPRW
• Language modeling
• “Senteval: An evaluation toolkit for universal sentence representations”, 2018 arXiv
• Object Detection, Segmentation
• “Faster r-cnn: Towards real-time object detection with region proposal networks”, 2015 NIPS
• “R-fcn: Object detection via region-based fully convolutional networks”, 2016 NIPS
• “Speed/accuracy trade-offs for modern convolutional object detectors”, 2017 CVPR
• “Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and
fully connected crfs”, 2017 TPAMI
5. Related Works
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 5
Transfer Learning with fine-tuning or frozen feature-based methods
• “Analyzing the performance of multilayer neural networks for object recognition”, 2014 ECCV
• “Return of the devil in the details: Delving deep into convolutional nets”, 2014 arXiv
• “Rich feature hierarchies for accurate object detection and semantic seg- mentation”, 2014 CVPR
• “How transferable are features in deep neural networks?”, 2014 NIPS
• “Factors of transferability for a generic convnet representation”, 2015 TPAMI
• “Bilinear cnn models for fine- grained visual recognition”, 2015 ICCV
• “What makes ImageNet good for transfer learning?”, 2016 arXiv
• “Best practices for fine-tuning visual classifiers to new domains”, 2016 ECCV
→ They show that fine-tuning outperforms frozen feature-based methods
6. Related Works
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 6
Adversarial robustness
• “Towards deep learning models resistant to adversarial attacks”, 2018 ICLR
• “Virtual adversarial training: a regularization method for supervised and semi-supervised learning”,
2018
• “Provably robust deep learning via adversarially trained smoothed classifier”, 2019 NeurIPS
• And many papers has studied the features learned by these robust networks and suggested that they
improve upon those learned by standard networks.
• On the other hand, prior studies have also identified theoretical and empirical tradeoffs between
standard accuracy and adversarial robustness.
7. Related Works
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 7
Adversarial robustness and Transfer learning
• “Adversarially robust transfer learning”, 2019 arXiv
• Transfer learning can increase downstream-task adversarial robustness
• “Adversarially-Trained Deep Nets Transfer Better”, 2020 arXiv
• Investigate the transfer performance of adversarially robust networks. → Very similar work!
• Authors study a larger set of downstream datasets and tasks and analyze the effects of model
accuracy, model width, and data resolution.
8. Experiments
Motivation: Fixed-Feature Transfer Learning
• Basically we use the source model as a feature extractor for the target dataset, the trains a simple (often
linear) model on the resulting features
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 8
Reference: Stanford cs231n lecture note
9. Experiments
How can we improve transfer learning?
• Prior works suggest that accuracy on the source dataset is a strong indicator of performance on
downstream tasks.
• Still, it is unclear if improving ImageNet accuracy is the only way to improve performance.
• After all, the behavior of fixed-feature transfer is governed by models’ learned representations, which
are not fully described by source-dataset accuracy.
• These representations are, in turn, controlled by the priors that we put on them during training
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 9
architectural components, loss functions, augmentations, etc.
10. Experiments
The adversarial robustness prior
• Adversarial robustness refers to a model’s invariance to small (often imperceptible) perturbations of its
inputs.
• Robustness is typically induced at training time by replacing the standard empirical risk minimization
objective with a robust optimization objective
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 10
11. Experiments
Should adversarial robustness help fixed-feature transfer?
• In fact, adversarially robust models are known to be significantly less accurate than their standard
counterparts.
• It suggest that using adversarially robust feature representations should hurt transfer performance.
• On the other hand, recent work has found that the feature representations of robust models carry
several advantages over those of standard models.
• For example, adversarially robust representations typically have better-behaved gradients and thus
facilitate regularization-free feature visualization
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 11
12. Experiments
Experiments – Fixed Feature Transfer Learning
• To resolve these two conflicting hypotheses (adversarially robust feature representations should hurt
transfer performance. vs. feature representations of robust models carry several advantages over
those of standard models.), use a test bed of 12 standard transfer learning datasets.
• Use four ResNet-based architecture (ResNet-18, 50, WideResNet-50-x2, 50-x4)
• The results indicate that robust networks consistently extract better features for transfer learning than
standard networks.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 12
13. Experiments
Experiments – Fixed Feature Transfer Learning
• To resolve these two conflicting hypotheses (adversarially robust feature representations should hurt
transfer performance. vs. feature representations of robust models carry several advantages over
those of standard models.), use a test bed of 12 standard transfer learning datasets.
• Use four ResNet-based architecture (ResNet-18, 50, WideResNet-50-x2, 50-x4)
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 13
14. Experiments
Experiments – Full-Network Fine Tuning
• A more expensive but often better-performing transfer learning method uses the pre-trained model as a
weight initialization rather than as a feature extractor.
• In other words, update all of the weights of the pre-trained model (via gradient descent) to minimize loss
on the target task.
• Many previous works find that for standard models, performance on full-network transfer learning is
highly correlated with performance on fixed-feature transfer learning.
• Hope that the findings of the last section (fixed-feature) also carry over to this setting (full-network).
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 14
15. Experiments
Experiments – Full-Network Fine Tuning
• Robust models match or improve on standard models in terms of transfer learning performance.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 15
16. Experiments
Experiments – Full-Network Fine Tuning
• Also, adversarially robust networks consistently outperform standard networks in Object Detection &
Instance Segmentation
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 16
17. Analysis & Discussion
4.1 ImageNet accuracy and transfer performance
• Take a closer look at the similarities and differences in transfer learning between robust networks and
standard networks.
• Hypothesis: robustness and accuracy have counteracting yet separate effects!
• That is, higher accuracy improves transfer learning for a fixed level of robustness, and higher
robustness improves transfer learning for a fixed level of accuracy
• The results (cf. Figure 5; similar results for full-network transfer in Appendix F) support this hypothesis.
• The previously observed linear relationship between accuracy and transfer performance is often violated
once robustness aspect comes into play.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 17
19. Analysis & Discussion
4.1 ImageNet accuracy and transfer performance
• In even more direct support of our hypothesis, find that when the robustness level is held fixed, the
accuracy- transfer correlation observed by prior works for standard models holds for robust models too.
• This findings also indicate that accuracy is not a sufficient measure of feature quality or versatility.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 19
20. Analysis & Discussion
4.2 Robust models improve with width
• Previous works find that although increasing network depth improves transfer performance, increasing
width hurts it.
• The results corroborate this trend for standard networks but indicate that it does not hold for robust
networks, at least in the regime of widths tested.
• As width increases, transfer performance plateaus and decreases for standard models, but continues to
steadily grow for robust models.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 20
Not always!!
21. Analysis & Discussion
4.2 Robust models improve with width
• Previous works find that although increasing network depth improves transfer performance, increasing
width hurts it.
• The results corroborate this trend for standard networks but indicate that it does not hold for robust
networks, at least in the regime of widths tested.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 21
22. Analysis & Discussion
4.3 Optimal robustness levels for downstream tasks
• Although the best robust models often outperform the best standard models, the optimal choice of
robustness parameter ε varies widely between datasets. For example, when transferring to CIFAR- 10
and CIFAR-100, the optimal ε values were 3.0 and 1.0, respectively.
• In contrast, smaller values of ε (smaller by an order of magnitude) tend to work better for the rest of the
datasets.
• One possible explanation for this variability in the optimal choice of ε might relate to dataset granularity.
• Although we lack a quantitative notion of granularity (in reality, features are not simply singular pixels),
we consider image resolution as a crude proxy.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 22
23. Analysis & Discussion
4.3 Optimal robustness levels for downstream tasks
• Since we scale target datasets to match ImageNet dimensions, each pixel in a low-resolution dataset
(e.g., CIFAR-10) image translates into several pixels in transfer, thus inflating dataset’s separability.
• Attempt to calibrate the granularities of the 12 image classification datasets used in this work, by first
downscaling all the images to the size of CIFAR-10 (32× 32), and then upscaling them to ImageNet size
once more.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 23
24. Analysis & Discussion
4.3 Optimal robustness levels for downstream tasks
• After controlling for original dataset dimension, the dataset’s epsilon vs. transfer accuracy curves all
behave almost identically to CIFAR-10 and CIFAR-100 ones. (Similar results for full-network transfer)
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 24
25. Analysis & Discussion
4.4 Comparing adversarial robustness to texture robustness
• Consider texture-invariant models, i.e., models trained on the texture-randomizing Stylized ImageNet
(SIN) dataset.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 25
26. Analysis & Discussion
4.4 Comparing adversarial robustness to texture robustness
• Transfer learning from adversarially robust models outperforms transfer learning from texture-invariant
models on all considered datasets.
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better? 26
Full-network
Fixed-feature
27. Conclusion
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better?
• Propose using adversarially robust models for transfer learning.
• Compare transfer learning performance of robust and standard models on a suite of 12
classification tasks, object detection, and instance segmentation.
• Find that adversarial robust neural networks consistently match or improve upon the
performance of their standard counterparts, despite having lower ImageNet accuracy.
• Take a closer look at the behavior of adversarially robust networks, and study the interplay
between ImageNet accuracy, model width, robustness, and transfer performance.
27
28. Conclusion
PR-290 | Do Adversarially Robust ImageNet Models Transfer Better?
• We can simply try this experiments! (https://github.com/Microsoft/robust-models-transfer)
28