• Publications
  • Influence
Transferable Clean-Label Poisoning Attacks on Deep Neural Nets
TLDR
We propose a new "polytope attack" in which poison images are designed to surround the targeted image in feature space. Expand
  • 52
  • 7
  • PDF
R-FCN-3000 at 30fps: Decoupling Detection and Classification
TLDR
We propose a modular approach to the largescale object detection problem that outperforms YOLO9000 by 18% while processing 30 images per second. Expand
  • 49
  • 7
  • PDF
Joint Human Detection and Head Pose Estimation via Multistream Networks for RGB-D Videos
TLDR
We propose a multistream multitask deep network for joint human detection and head pose estimation in RGB-D videos. Expand
  • 29
  • 3
Learning From Noisy Anchors for One-Stage Object Detection
TLDR
In this paper, we propose to mitigate noise incurred by imperfect label assignment such that the contributions of anchors are dynamically determined by a carefully constructed cleanliness score associated with each anchor. Expand
  • 14
  • 1
  • PDF
Multi-glimpse LSTM with color-depth feature fusion for human detection
TLDR
We propose a new Multi-Glimpse LSTM (MG-LSTM) network, in which multi-scale contextual information is sequentially integrated to promote the human detection performance. Expand
  • 10
  • 1
  • PDF
A Dynamic Frame Selection Framework for Fast Video Recognition.
TLDR
We introduce AdaFrame, a conditional computation framework that adaptively selects relevant frames on a per-input basis for fast video recognition using a Long Short-Term Memory augmented with a global memory. Expand
  • 2
  • 1
An Analysis of Pre-Training on Object Detection
TLDR
We provide a detailed analysis of convolutional neural networks which are pre-trained on the task of object detection on large datasets like OpenImagesV4, ImageNet Localization and COCO. Expand
  • 18
  • PDF
Improving the Tightness of Convex Relaxation Bounds for Training Certifiably Robust Classifiers
TLDR
We propose two regularizers that can be used to train neural networks that yield tighter convex relaxation bounds for robustness. Expand
  • 3
  • PDF
2D or not 2D? Adaptive 3D Convolution Selection for Efficient Video Recognition
TLDR
We introduce Ada3D, a conditional computation framework that learns instance-specific 3D usage policies to determine frames and convolution layers to be used in a 3D network. Expand
  • 1
  • PDF
...
1
2
...