• Publications
  • Influence
Diverse Image-to-Image Translation via Disentangled Representations
TLDR
We present an approach based on disentangled representation for producing diverse outputs without paired training images. Expand
  • 480
  • 105
  • PDF
Unsupervised Representation Learning by Sorting Sequences
TLDR
We present an unsupervised representation learning approach using videos without semantic labels and train a convolutional neural network to sort the shuffled sequences. Expand
  • 230
  • 25
  • PDF
Mode Seeking Generative Adversarial Networks for Diverse Image Synthesis
TLDR
We propose a simple yet effective regularization term that can be applied to cGANs for various tasks to alleviate the mode collapse problem. Expand
  • 91
  • 20
  • PDF
Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformation
TLDR
We use feature-wise transformation layers for augmenting the image features using affine transforms to simulate various feature distributions under different domains in the training stage. Expand
  • 43
  • 10
  • PDF
Soft-Segmentation Guided Object Motion Deblurring
TLDR
We propose an efficient algorithm to jointly estimate object segmentation and camera motion where each layer can be deblurred well under the guidance of the soft-segmentation. Expand
  • 55
  • 7
  • PDF
Bio-Inspired Proximity Discovery and Synchronization for D2D Communications
TLDR
In this paper, a distributed mechanism for application-aware proximity services (ProSe) in device to device communications is presented. Expand
  • 47
  • 4
  • PDF
NTUplace4dr: A Detailed-Routing-Driven Placer for Mixed-Size Circuit Designs With Technology and Region Constraints
TLDR
We present a detailed-routability-driven analytical placement algorithm for modern mixed-size designs with technology and region constraints. Expand
  • 15
  • 4
Cleavage-site specificity of prolyl endopeptidase FAP investigated with a full-length protein substrate.
Fibroblast activation protein (FAP) is a prolyl-cleaving endopeptidase proposed as an anti-cancer drug target. It is necessary to define its cleavage-site specificity to facilitate the identificationExpand
  • 20
  • 1
  • PDF
Controllable Image Synthesis via SegVAE
TLDR
We propose SegVAE, a VAE-based framework that can generate semantic maps in an iterative manner using conditional variational autoencoder. Expand
  • 4
  • 1
  • PDF
Dancing to Music
TLDR
We propose a synthesis-by-analysis learning framework to generate dance from music through a decomposition-to-composition framework. Expand
  • 3
  • 1
  • PDF