The Art of Deep Learning

General CNN papers

• Spatial Transformer Networks link -- 2016
• Non-local Neural Networks link -- 2017
• Network in Network (NIN): link
• Striving for simplicity: The all convolutional net link
• Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition (SPP) link
• Deeply Supervised Nets link -- 2014

Representation Learning:

• Representation Learning: A Review and New Perspectives link
• Regularized Auto-Encoders Estimate Local Statistics link
• Better Mixing via Deep Representations link
• Spatial Transformer Networks link

CNN-based Face Recognition (Identification/Verification):

• The Do’s and Don’ts for CNN-based Face Verification link

• A summary of deep models for face recognition pdf

• Labeled Faces in the Wild: A Database for Studying Face Recognition in Unconstrained Environments pdf (describes LFW dataset and the unseen pair matching problem ) (independent decision for all pairs)

• Unconstrained face verification using deep cnn features link

• Frankenstein: Learning deep face representations using small data link

• Triplet probabilistic embedding for face verification and clustering link

• Facenet: A unified embedding for face recognition and clustering link

• Deepid3: Face recognition with very deep neural networks link

• Hybrid deep learning for face verification

• Deepface: Closing the gap to human-level performance in face verification pdf

• Deep face recognition pdf

• A Discriminative Feature Learning Approach for Deep Face Recognition pdf

• Joint face detection and alignment using multitask cascaded convolutional networks link

Detection and Segmentation:

• Panoptic Feature Pyramid Networks (Panoptic FPN) link
• You Only Look Once: Unified, Real-Time Object Detection link
• Mask R-CNN link
• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks link
• Fast R-CNN link

Adversarial Attacks:

• There Is No Free Lunch In Adversarial Robustness (But There Are Unexpected Benefits) link
• (Szegedy; L-BFGS) Intriguing properties of neural networks link
• (FGS) Explaining and harnessing adversarial examples link
• Adversarial Manipulation of Deep Representations link
• ADVERSARIAL EXAMPLES IN THE PHYSICAL WORLD (iterative methods) link
• Adversarial machine learning at scale link
• (hot/cold) Adversarial diversity and hard positive generation link
• Assessing Threat of Adversarial Examples on Deep Neural Network link
• Simple Black-Box Adversarial Perturbations for Deep Networks link
• Are Facial Attributes Adversarially Robust? link

Defense Against Adversarial Attacks (Adversarial Robustness)

• Fortified Networks: Improving the Robustness of Deep Networks by Modeling the Manifold of Hidden Representations link
• Defense-GAN: Protecting Classifiers Against Adversarial Attacks Using Generative Models lin
• Ensemble Methods as a Defense to Adversarial Perturbations Against Deep Neural Networks link
• Adversarial robustness: Softmax versus Openmax link
• Towards Robust Deep Neural Networks with BANG link
• Ensemble Adversarial Training: Attacks and Defenses link
• Robustness of classifiers: from adversarial to random noise link
• Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulation link
• Lower bounds on the robustness to adversarial perturbations link
• Improving the Robustness of Deep Neural Networks via Stability Training link
• Towards Deep Neural Network Architectures Robust to Adversarial Examples link
• Foveation-based Mechanisms Alleviate Adversarial Examples link
• Adversarial Logit Pairing link
• Towards deep learning models resistant to adversarial attacks link

Generative Models

• Auto-Encoding Variational Bayes (aka VAE) link
• Generative Adversarial Networks link
• Adversarial Autoencoders link
• Generative Probabilistic Novelty Detection with Adversarial Autoencoders link

GAN applicaitons:

• WarpGAN: Automatic Caricature Generation link