• Publications
  • Influence
Understanding deep image representations by inverting them
Image representations, from SIFT and Bag of Visual Words to Convolutional Neural Networks (CNNs), are a crucial component of almost any image understanding system. Nevertheless, our understanding ofExpand
  • 1,160
  • 72
  • PDF
Visualizing Deep Convolutional Neural Networks Using Natural Pre-images
TLDR
We study several representations, both shallow and deep, by a number of complementary visualization techniques. Expand
  • 298
  • 24
  • PDF
Object-Centric Learning with Slot Attention
TLDR
We present the Slot Attention module, an architectural component that interfaces with perceptual representations such as the output of a convolutional neural network and produces a set of task-dependent abstract representations which we call slots. Expand
  • 18
  • 6
  • PDF
Salient Deconvolutional Networks
TLDR
In this paper, we introduce a family of reversed networks that generalizes and relates deconvolution, backpropagation and network saliency. Expand
  • 72
  • 3
  • PDF
Cross Pixel Optical Flow Similarity for Self-Supervised Learning
TLDR
We propose a novel method for learning convolutional neural image representations without manual supervision. Expand
  • 29
  • 1
  • PDF
Heterogeneous UGV-MAV exploration using integer programming
TLDR
This paper presents a novel exploration strategy for coordinated exploration between unmanned ground vehicles (UGV) and micro-air vehicles (MAV). Expand
  • 20
  • PDF
Self-Supervised Learning of Video-Induced Visual Invariances
TLDR
We propose a general framework for self-supervised learning of transferable visual representations based on Video-Induced Visual Invariances (VIVI). Expand
  • 20
  • PDF
ResearchDoom and CocoDoom: Learning Computer Vision with Games
TLDR
In this short note we introduce ResearchDoom, an implementation of the Doom first-person shooter that can extract detailed metadata from the game that can be used to train and evaluate a variety of computer vision algorithms. Expand
  • 14
  • PDF
UGV-MAV Collaboration for Augmented 2D Maps
TLDR
We propose a novel collaborative system, consisting of an unmanned ground vehicle and a micro aerial vehicle, which is used to create augmented 2D maps using the distinct sensing capabilities of these two robots. Expand
  • 3
Self-supervised Segmentation by Grouping Optical-Flow
TLDR
We propose to self-supervise a convolutional neural network operating on images by learning to group image pixels in such a way that their collective motion is “coherent”. Expand
  • 2
  • PDF