PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
arXiv:1608.08021
Abstract
This paper presents how we can achieve the state-of-the-art accuracy in multi-category object detection task while minimizing the computational cost by adapting and combining recent technical innovations. Following the common pipeline of "CNN feature extraction + region proposal + RoI classification", we mainly redesign the feature extraction part, since region proposal part is not computationally expensive and classification part can be efficiently compressed with common techniques like truncated SVD. Our design principle is "less channels with more layers" and adoption of some building blocks including concatenated ReLU, Inception, and HyperNet. The designed network is deep and thin and trained with the help of batch normalization, residual connections, and learning rate scheduling based on plateau detection. We obtained solid results on well-known object detection benchmarks: 83.8% mAP (mean average precision) on VOC2007 and 82.5% mAP on VOC2012 (2nd place), while taking only 750ms/image on Intel i7-6700K CPU with a single core and 46ms/image on NVIDIA Titan X GPU. Theoretically, our network requires only 12.3% of the computational cost compared to ResNet-101, the winner on VOC2012.
Full details about "PVANet 9.0" in the VOC2012 leaderboard (https://goo.gl/DuQBku). The test codes are available at https://github.com/sanghoon/pva-faster-rcnn
Cited by in corpus (12)
- ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices
- PixelLink: Detecting Scene Text via Instance Segmentation
- Real-time Scene Text Detection with Differentiable Binarization
- Text Detection and Recognition in the Wild: A Review
- Embedded Real-Time Fall Detection Using Deep Learning For Elderly Care
- Training Bit Fully Convolutional Network for Fast Semantic Segmentation
- Towards High Performance Video Object Detection
- Synthetic Data Generation and Adaption for Object Detection in Smart Vending Machines
- Road images augmentation with synthetic traffic signs using neural networks
- Integrating Deep Learning and Augmented Reality to Enhance Situational Awareness in Firefighting Environments
- Efficient Scene Text Detection with Textual Attention Tower
- Continuity-Discrimination Convolutional Neural Network for Visual Object Tracking