Scalable learning for object detection with GPU hardware

We consider the problem of robotic object detection of such objects as mugs, cups, and staplers in indoor environments. While object detection has made significant progress in recent years, many current approaches involve extremely complex algorithms, and are prohibitively slow when applied to large scale robotic settings. In this paper, we describe an object detection system that is designed to scale gracefully to large data sets and leverages upward trends in computational power (as exemplified by Graphics Processing Unit (GPU) technology) and memory. We show that our GPU-based detector is up to 90 times faster than a well-optimized software version and can be easily trained on millions of examples. Using inexpensive off-the-shelf hardware, it can recognize multiple object types reliably in just a few seconds per frame.

Authors

• Adam Coates 0002
• Paul Baumstarck
• Quoc V. Le
• Andrew Y. Ng

Keywords

• Object detection
• Hardware
• Detectors
• Intelligent robots
• Graphics
• Robot sensing systems
• Moore's Law
• Clocks
• USA Councils
• Indoor environments
• Scalable
• Graphics Processing Unit
• Graphics Processing Unit Hardware
• Large Datasets
• Power Calculation
• Progress In Recent Years
• Histogram
• Training Set
• Decision Tree
• Parallelization
• Grayscale Images
• Object Classification
• Target Object
• Channel Images
• Training Examples
• Feature Calculation
• Single Machine
• Large Training Set
• Moore’s Law
• Memory Bandwidth
• Normalized Cross-correlation
• Object Detection Approaches
• Memory Transfer
• Perception Of The Robot
• Background Pattern
• Image Retrieval
• Gini Coefficient