PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection

By Kye-Hyeon Kim, Yeongjae Cheon, Sanghoon Hong, Byungseok Roh, Minje Park. 2016

218 citations recorded by the source corpus.

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.

Subfields: Computer Vision and Pattern Recognition, Industrial and Manufacturing Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Industrial and Manufacturing Engineering

Citations in the corpus

Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors (2016, background): “Since appearing in 2015, Faster R-CNN has been particularly influential, and has led to a number of follow-up works [2, 34, 33, 45, 13, 5, 19, 44, 23, 46] (including SSD and R-FCN).”
ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices (2017, background): “Recently, Kim et al. propose a lightweight network structure named PVANET [20] which adopts Inception units.”
Learning Efficient Object Detection Models with Knowledge Distillation (2017, background): “Thus, to achieve faster speeds, some prior works explore new structures such as fully convolutional networks, or lightweight models with fewer channels and small ﬁlters [22,25].”
Deep Learning for Generic Object Detection: A Survey (2018, methodology): “n the COCO detection challenge 2015. HyperNet [135] RPN VGG16 Faster RCNN 76:3 (07+12) 71:4 (07T+12) CVPR16 Use features from multiple layers for both region proposal and region classiﬁcation. PVANet [132] RPN PVANet Faster RCNN 84.9 (07+12+CO) 84.2 (07T+12+CO) NIPSW16 Deep but lightweight; Combine ”
EAST: An Efficient and Accurate Scene Text Detector (2017, methodology): “Even the best performing model Ours+PVANET2x runs at a speed of 13.2 FPS.”
Scene Text Detection and Recognition: The Deep Learning Era (2018, methodology): “Instead of VGG [129], EAST uses PVANet [67] as its basenetwork, which strikes a good balance between efficiency and accuracy in the ImageNet competition.”
Real-time Scene Text Detection with Differentiable Binarization (2019, methodology): “EAST (Zhou et al. 2017) proposed to apply PVANet (Kim et al. 2016) to improve its speed.”
PixelLink: Detecting Scene Text via Instance Segmentation (2018, background): “PVANET2x is a modified version of PVANET (Kim et al. 2016), by doubling the channels.”
Pelee: A Real-Time Object Detection System on Mobile Devices (2018, methodology): “[20] widely uses inception block in PVANet, which is a high accuracy real time object detection system.”
TextBoxes++: A Single-Shot Oriented Scene Text Detector (2018, unknown): “To improve the performance, the authors double the number of channels of PVANet, which results in a runtime at 13.2 fps. TextBoxes++ has a similar runtime but with a 3.5 percents performance improvement.”
Object Detection in 20 Years: A Survey (2019, methodology): “This idea has been recently used in object detection [200].”
Synthetic Data for Deep Learning (2019, result): “report significantly improved results when using hybrid datasets of real and synthetic data for all three tested object detection architectures: PVANET [320], SSD [372], and YOLOv3 [482].”
Recent Advances in Deep Learning for Object Detection (2019, methodology): “PVANet [104] was proposed as a new network structure with CReLu [172] layer to reduce non-linear computation and accelerated inference speed.”
Receptive Field Block Net for Accurate and Fast Object Detection (2017, background): “Its variants [35, 33, 17] achieve competitive results in object detection and classification tasks.”