• Publications
  • Influence
Axial-DeepLab: Stand-Alone Axial-Attention for Panoptic Segmentation
TLDR
We propose a position-sensitive self-attention layer, a novel building block that one could stack to form axial-att attention models for image classification and dense prediction. Expand
  • 52
  • 9
  • PDF
Weight Standardization
TLDR
In this paper, we propose Weight Standardization (WS), which smooths the loss landscape by standardizing the weights in convolutional layers. Expand
  • 36
  • 6
Semantic-Aware Knowledge Preservation for Zero-Shot Sketch-Based Image Retrieval
TLDR
Sketch-based image retrieval (SBIR) is widely recognized as an important vision problem which implies a wide range of real-world applications. Expand
  • 21
  • 5
  • PDF
Image-based visual adaptive tracking control of nonholonomic mobile robots
TLDR
A novel image-based visual approach for controlling nonholonomic mobile robots to track a moving target is presented. Expand
  • 50
  • 1
Single 2D pressure footprint based person identification
TLDR
We propose an automatic footprint based person identification method using a single bare or socked footprint, which differs from the existing bare footprint based methods. Expand
  • 3
  • 1
Combining Compositional Models and Deep Networks For Robust Object Classification under Occlusion
TLDR
We combine DC-NNs and compositional object models to retain the best of both approaches: a discriminative model that is robust to partial occlusion and mask attacks. Expand
  • 17
  • 1
  • PDF
A Slicing Algorithm of Concurrency Modeling Based on Petri Nets
  • 10
  • 1
ELASTIC: Improving CNNs with Instance Specific Scaling Policies
TLDR
We introduce ELASTIC, a simple, efficient and yet very effective approach to learn instance-specific scale policy from data. Expand
  • 5
  • 1
Micro-Batch Training with Batch-Channel Normalization and Weight Standardization
TLDR
We propose Weight Standardization (WS) and Batch-Channel Normalization (BCN) to bring two success factors of BN into micro-batch training: 1) smoothing effects on the loss landscape and 2) the ability to avoid harmful elimination singularities along the training trajectory. Expand
  • 36
  • PDF
...
1
2
3
4
...