• Publications
  • Influence
Aggregated Residual Transformations for Deep Neural Networks
TLDR
We present a simple, highly modularized network architecture for image classification. Expand
  • 2,887
  • 555
  • PDF
Holistically-Nested Edge Detection
TLDR
We develop a new edge detection algorithm that addresses two important issues in this long-standing vision problem: (1) holistic image training and prediction; and (2) multi-scale and multi-level feature learning. Expand
  • 1,472
  • 307
  • PDF
Momentum Contrast for Unsupervised Visual Representation Learning
TLDR
We present Momentum Contrast (MoCo) for unsupervised visual representation learning. Expand
  • 453
  • 152
  • PDF
Rethinking Spatiotemporal Feature Learning: Speed-Accuracy Trade-offs in Video Classification
TLDR
We show that it is possible to replace many of the 3D convolutions at the lowest layers of the network (the ones closest to the pixels), and use 2D convolution for the higher layers. Expand
  • 312
  • 74
  • PDF
Deeply-Supervised Nets
TLDR
Our proposed deeply-supervised nets (DSN) method simultaneously minimizes classification error while making the learning process of hidden layers direct and transparent. Expand
  • 1,103
  • 72
  • PDF
Exploring Randomly Wired Neural Networks for Image Recognition
TLDR
We explore a more diverse set of connectivity patterns through the lens of randomly wired networks that are sampled from stochastic network generators. Expand
  • 146
  • 24
  • PDF
Rethinking Spatiotemporal Feature Learning For Video Understanding
TLDR
In this paper we study 3D convolutional networks for video understanding tasks. Expand
  • 114
  • 22
Decoupling Representation and Classifier for Long-Tailed Recognition
TLDR
The long-tail distribution of the visual world poses great challenges for deep learning based classification models on how to handle the class imbalance problem. Expand
  • 38
  • 14
  • PDF
Hyper-class augmented and regularized deep learning for fine-grained image classification
TLDR
In this paper, we propose a principled framework to explicitly tackle the challenges of learning a deep CNN for fine-grained image classification with small training data. Expand
  • 123
  • 8
  • PDF
Attentional ShapeContextNet for Point Cloud Recognition
TLDR
ShapeContextNet is an end-to-end model that can be applied to the general point cloud classification and segmentation problems. Expand
  • 86
  • 8
  • PDF