Computer Vision

General

• Densely Connected Convolutional Networks
• Active Convolution: Learning the Shape of Convolution for Image Classification
• DiracNets: Training Very Deep Neural Networks Without Skip-Connections

Object Detection

• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

Adversarial / Variational

• GAN: Generative Adversarial Networks
• DCGAN: Semantic Image Inpainting with Perceptual and Contextual Losses
• BEGAN: Boundary Equilibrium Generative Adversarial Networks
• Improved Training of Wasserstein GANs
• Progressive Growing of GANs for Improved Quality, Stability, and Variation

Visual QA

• Visual Question Answering with Memory-Augmented Networks