Adaptive Object Detection Using Adjacency and Zoom Prediction
arXiv:1512.07711
Abstract
State-of-the-art object detection systems rely on an accurate set of region proposals. Several recent methods use a neural network architecture to hypothesize promising object locations. While these approaches are computationally efficient, they rely on fixed image regions as anchors for predictions. In this paper we propose to use a search strategy that adaptively directs computational resources to sub-regions likely to contain objects. Compared to methods based on fixed anchor locations, our approach naturally adapts to cases where object instances are sparse and small. Our approach is comparable in terms of accuracy to the state-of-the-art Faster R-CNN approach while using two orders of magnitude fewer anchors on average. Code is publicly available.
Accepted to CVPR 2016
References in corpus (7)
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
- Caffe: Convolutional Architecture for Fast Feature Embedding
- Going Deeper with Convolutions
- Inside-Outside Net: Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks
- Computational Baby Learning
- Efficient Object Detection for High Resolution Images
Cited by in corpus (6)
- Hierarchical Object Detection with Deep Reinforcement Learning
- Tree-Structured Reinforcement Learning for Sequential Object Localization
- Dynamic Zoom-in Network for Fast Object Detection in Large Images
- Active Object Localization in Visual Situations
- Spatial Memory for Context Reasoning in Object Detection
- Fast On-Line Kernel Density Estimation for Active Object Localization