Robust Object detection for tiny and dense targets in VHR Aerial Images
|
|
- Griselda Stephens
- 5 years ago
- Views:
Transcription
1 Robust Object detection for tiny and dense targets in VHR Aerial Images Haining Xie 1,Tian Wang 1,Meina Qiao 1,Mengyi Zhang 2,Guangcun Shan 1,Hichem Snoussi 3 1 School of Automation Science and Electrical Engineering, Beihang University, Beijing, China; 2 College of Electrical Engineering and Control Science, Nanjing Tech University, Nanjing, China; 3 Institute Charles Delaunay-LM2S-UMR STMR 6279 CNRS, University of Technology of Troyes, France. xiehaining@buaa.edu.cn, wangtian@buaa.edu.cn, meinaqiao@buaa.edu.cn, myzhang@njtech.edu.cn, gcshan@buaa.edu.cn, hichem.snoussi@utt.fr Abstract Object detection in Very High Resolution (VHR) optical remote sensing images is a challenged work for objects are usually dense and tiny. With random orientation, various backgrounds as well as unpredictable noise make traditional image processing methods perform badly. In this paper, we propose using state-of-art Region-based fully convolutional networks to solve object detection tasks in aerial images. To make the whole system efficient we choose to utilize position-sensitive score maps which not only fully take advantage of the convolutional feature maps but also achieve a balance between translation-variance in object detection and translation-invariance in classification. In addition, with 101-layer Residual networks as feature extractors, we achieve a satisfying result which is low time consuming and shows percent and percent precision respectively on two datasets. Index Terms Object Detection, Convolutional Neural Network, NWPU VHR-10 dataset, Pri-SDL dataset I. INTRODUCTION With excellent abstraction ability of deep convolutional neural network, related measures have been tried in remote sensing images to help solve object detection and classification issues. Also, in the past few years, a nearly complete framework of object detection for common scenes has been developed. First of the following parts will generally summarize the major progress of object detection for common scenes. Then the second part will focus on the latest researches on object detection in aerial images. A. Object detection models in common images Different convolutional neural networks like Alex net[1], VGG net[2], Google net[3] have been proposed and achieved an outstanding result in classification which was even better than human expert. An original object detection model of a prevalent family[4] [5] [6], RCNN[4] utilized those excellent classification neural netwoks as fundamental feature extractors. Even though it made a great progress in object detection work, it computed slowly due to repetitive computational work and was hard to be trained for its complicated components. Fixed image size was another annoying limit caused by fully connected layer. In this version, selective search[7] has been chosen to propose region of interest(roi) among several methods proposed recently like objectness[8], multiscale combinatorial grouping[9], category-independent object proposals[10]. The successors of RCNN are SPP net[11] and Fast RCNN[5]. SPP net suggested spatial pyramid pooling to free the limit of fixed image size while Fast RCNN proposed ROI max pooling.generally, ROI max pooling layer is a particular case of spatial pooling layer. Fast RCNN used a shared convolutional layer of the whole image as its feature map, which reduced superfluous computation and trained 9 faster than RCNN. With all these improvements, Fast RCNN became a mearly end-to-end neural network, which meant it can be trained more easily. The latest version is Faster RCNN[6] which suggested regional proposal network which was a little fully convolutional network merged with the whole network to propose ROI more efficiently. Another creative thought was anchors which can be thought as a regression method of spatial pyramid pooling[6]. All these composed a unified convolutional network which can be trained and used efficiently. In this paper, we propose using Regionbased fully convolutional network which is an improvement of the former prevalent family inspired by recent semantic segmentation tasks[12] for object detection in VHR aerial images. It proposes position-sensitive score maps[13] to make a compromise of basic conflict in object detection. B. Object detection in VHR optical remote sensing images According to past researches, considerable traditional methods have been developed to detect tiny and dense objects in aerial images. Several hand-craft features like HOG[14],Haarlike[15],LBP[16] were used to represent objects. These approaches were easy to understand and stayed at a primitive stage and usually needed to change according to specific environments. Also they did not present satisfying results and did cost a lot of computational resources because multiple oriented channels were usually used to merge results. Nowadays convolutional layers and pooling layers show robust abstraction ability, however, previous experts tried hard to manually find out invariant features and specific transformation. Then latest researches used basic convolutional neural network which showed robust abstraction ability in object detection. The paper[17] used AlexNet[1] as their features extractor and conducted a coarse-location-fine-classification pipeline which was a particular adaptation to usually utilized
2 Fig. 1: Basic framework of Region-based Fully Convolutional Network. At first, images are convolved by ResNet-101 to generate feature maps. Then feature maps are convolved by two seperate networks, which are RPN(Region Proposal Network) and position-sensitive score maps. By pooling the ROI proposed by RPN, k k bins are used to vote. Selective Research method. And the paper[18] nearly chose the same settings as the former one but proposed a new rotation-invariant layer which just optimized the multinomial logistic regression objective. However, they paid attention to the invariant of rotation which was actually solved by common convolutional neural network. Then several neural network methods were introduced mainly to solve classification problems. We find out that all these methods used in aerial images are primitive and do cost a lot of computational resources. Their methods were not end-to-end, so it was complex and difficult when training. Also the particular differences between aerial images and images of common scenes were not fully discussed which might lead to a biased way. C. Proposal At first a comparison between SSD[19] and RFCN[13] was conducted in aerial images,and RFCN was chosen because of its accuracy and expandability. We propose using Regionbased Fully Convolutional Networks[13] as an elegant and end-to-end framework and suggest different parameters fine tuning methods for various conditions. After a comprehensive comparative analysis, we achieve a more accurate result which can adapt to scenes with dense and tiny objects. II. OUR APPROACH Figure 1 illustrates the basic framework of our structure. It contains two major parts: a residual neural network used as feature extractor and a popular two-stage object detection pipeline following. The object detection consists of two subsections: one is to propose ROI(region of interest) while the other is to classify ROI. A. Data Preparation A large dataset is usually necessary to train a neural network and several data argumentation methods like mirroring,rotating or scaling the image are usually used. We propose using a pretrained model(firstly we use a model trained on a dataset of pets) can significantly reduce the needs of extra data, however a dataset with only 100 images can also perform well in our test. B. Residual Neural Network As the backbone architecture, ResNet-101[20] is the basis of R-FCN, which is chosen because it performs excellent in ImageNet Classification competitions. And like GooLeNets[3], ResNet-101 is by design fully convolutional. The fully convolutional structure, instead of fully connected structure, gets rid of the limit of fixed image size which means it does not need to face the annoying image resizing issues. There are 100 convolutional layers in ResNet-101 while global average pooling, instead of usually used max pooling layers, and C-class fc layers are used to follow[20]. (C represents the number of classes) However, as the successor of the prevalent R-CNN family,only convolutional layers are kept to generate feature maps. To reduce the computational time in training, we take advantage of the transfer learning which shows even an unrelated pretrained model can be better than randomly initialized parameters. Our model is pre-trained on COCO dataset. Originally the ResNet ends up with a d convolutional block, however, to reduce the dimension, a 1024-d 1 1 convolutional layer which is randomly initialized is attached at the end. Then a (C +1) k-channel layer will be applied to generate position-sensitive score maps. C. Score maps and pooling layers Every ROI rectangle is divided into k kbins to encode position information. Fig 1 shows this structure and set k = 3.When pooling, we can take advantage of the following formula. r c (i, j θ) = (x,y) bin(i,j) z i,j,c (x + x 0,y+ y 0 θ)/n, (1) In the formula, r represents the response of the (i,j)-th bin for the c-th class after pooling. D. Adaption to dense and tiny objects This model performs well in images with common scenes where objects usually occupy large areas,however, in VHR aerial images,objects like planes or cars are tiny and distribute randomly. Usually objects perform a dense distribution which makes this task challenging. We propose that the size of the input feature maps and the size of the anchors need to be fine tuned to adapt to VHR aerial images especially for dense and tiny targets. III. EXPERIMENTS We evaluate the model on two publicly available plane datasets provided respectively by Pri-SDL dataset [17] and NWPU VHR-10 dataset [18]. The first dataset is divided into two subsets.one contains 500 images which are for training while the other contains 100 images used for evaluation. The second dataset is fully used to evaluate.
3 Fig. 2: Precision-versus-recall curves A. Details of settings Several parameters are set manually before training according to experience. Because we use the transfer learning method, a pre-trained model on non-relative dataset, we find that setting the initial learning rate at performs well. Anchors scale and Aspect ratio are chosen to use default values. Height stride and width stride are both set to 16.Using GTX 1060 with 6GB memory, 2 to 3 hours were used to train.the following experiments are all using this trained model. B. Performence on Pri-SDL dataset According to [17], there are 600 images with 3210 plane samples in all, which are collected from Google Earth. The evaluation standard is the same as the PASCAL VOC object detection evaluation protocol[21], which rules that the detected bounding boxes are considered to be matched with the ground truth when overlapping area is larger than 50%..Fig 2 shows the Precision-versus-recall curves compared with the results given by [17]. Setting the recall rate at 0.9, we compare precision and present the result in Table I. Feature ACF FC6+FC7 POOL5+FC6+FC7 R-FCN Plane TABLE I: Performance when setting recall at 0.9 It can be seen that both DCNN and R-FCN outperform the ACF method and R-FCN is more robust that DCNN which keeps precision rate when recall achieves Also it can be inferred that R-FCN is time saving because DCNN does not share convolutional features and performs repeated computation.fig 3shows some final examples. C. Evaluation on NWPU VHR-10 dataset Totally there are 80 images with 757 airplane samples in NWPU VHR-10 dataset,the spatial resolution of which range from 0.5 to 2m. We use this whole dataset only for evaluation. The former R-FCN model trained on [17] dataset has been used with no more new adaption, which can also prove the robustness.precision-recall curve and Average Precision,both Fig. 3: The final results of Pri-SDL dataset. are standard and wieldly used, are performed to evaluate and compare. Then the average running time per image is compared to present the efficiency of R-FCN in aerial images. BoW SSCBoW FDDL COPD Transferred CNN RICNN R-FCN TABLE II: Preformance Comparisons of AP Values Fig 5 show that RICNN with fine-tuning performs significantly better than previous methods like COPD.However, it can be seen that the Precision-Recall curve of R-FCN is above the curve of RICNN and keeps precision rate of when recall is at This shows that R-FCN is more robust than RICNN in detecting airplanes in VHR aerial images. Fig 6 shows the results generated by our trained model. Table.II compares the AP values of different algorithms while Fig. 4 presents the average running time per image. All
4 Fig. 4: Time consuming of different algorithms 3UHFLVLRQ YHUVXV UHFDOO FXUYHV RI SODQH GHWHFWLRQ 5 )&1 5,&11 &23' 3UHFLVLRQ 5HFDOO Fig. 5: Precision-Recall curve these show that R-FCN used in detecting airplanes in VHR aerial images is not only more robust and accurate,but also more time-saving. D. Dense and tiny targets The challenging part of object detection in VHR aerial images is that it usually has dense and tiny detection targets than common scene.but we find out that our approach can adapt these scenes well after fine tune the relative parameters. Though R-FCN gets rid of the limit of fixed image size, it can also be affected by the proportion between the target and the whole scene. By fine-tuning the size of the feature extraction maps and anchors, the R-FCN model shows well in scenes with even more dense and tiny targets. Fig.7show some examples. IV. C ONCLUSION In this paper, we propose using the object detection model, R-FCN, in VHR aerial images. We prove that pre-trained models from even non-ralative dataset may reduce training time dramatically. After comparing with other detection algorithms used in aerial images on different datasets, the results show that R-FCN is more robust and gain a higher AP score. Also we fine-tuning the size of feature extraction maps and anchors Fig. 6: The results of detecting planes in NWPU VHR-10 dataset
5 [14] N. Dalal and B. Triggs, Histograms of oriented gradients for human detection, vol. 1, pp , [15] P. Viola and M. Jones, Rapid object detection using a boosted cascade of simple features, vol. 1, pp , [16] T. Ojala, M. Pietikainen, and T. Maenpaa, Multiresolution gray-scale and rotation invariant texture classification with local binary patterns, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 24, no. 7, pp , [17] H. Q. Zhu, X. Chen, W. Dai, K. Fu, Q. Ye, and J. Jiao, Orientation robust object detection in aerial images using deep convolutional neural network, pp , [18] G. Cheng, P. Zhou, and J. Han, Learning rotation-invariant convolutional neural networks for object detection in vhr optical remote sensing images, IEEE Transactions on Geoscience and Remote Sensing, vol. 54, no. 12, pp , [19] W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C. Fu, and A. C. Berg, Ssd: Single shot multibox detector, pp , [20] K. He, X. Zhang, S. Ren, and J. Sun, Deep residual learning for image recognition, pp , [21] M. Everingham, S. M. Eslami, L. Van Gool, C. K. I. Williams, J. Winn, and A. Zisserman, The pascal visual object classes challenge: A retrospective, International Journal of Computer Vision, vol. 111, no. 1, pp , Fig. 7: The results of detecting dense and tiny planes. when facing more dense and tiny targets, which also obtains a satisfying results. REFERENCES [1] A. Krizhevsky, I. Sutskever, and G. E. Hinton, Imagenet classification with deep convolutional neural networks, pp , [2] K. Simonyan and A. Zisserman, Very deep convolutional networks for large-scale image recognition, [3] C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, and A. Rabinovich, Going deeper with convolutions, pp. 1 9, [4] R. Girshick, J. Donahue, T. Darrell, and J. Malik, Rich feature hierarchies for accurate object detection and semantic segmentation, pp , [5] R. Girshick, Fast r-cnn, pp , [6] S. Ren, K. He, R. Girshick, and J. Sun, Faster r-cnn: Towards real-time object detection with region proposal networks, IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 1 1, [7] J. Uijlings, K. E. Sande, T. Gevers, and A. Smeulders, Selective search for object recognition, International Journal of Computer Vision, vol. 104, no. 2, pp , [8] B. Alexe, T. Deselaers, and V. Ferrari, Measuring the objectness of image windows, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 34, no. 11, pp , [9] P. Arbelaez, J. Ponttuset, J. Barron, F. Marques, and J. Malik, Multiscale combinatorial grouping, pp , [10] I. Endres and D. Hoiem, Category independent object proposals, pp , [11] K. He, X. Zhang, S. Ren, and J. Sun, Spatial pyramid pooling in deep convolutional networks for visual recognition, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37, no. 9, pp , [12] J. Long, E. Shelhamer, and T. Darrell, Fully convolutional networks for semantic segmentation, pp , [13] J. Dai, Y. Li, K. He, and J. Sun, R-fcn: Object detection via regionbased fully convolutional networks, pp , 2016.
Deep learning for object detection. Slides from Svetlana Lazebnik and many others
Deep learning for object detection Slides from Svetlana Lazebnik and many others Recent developments in object detection 80% PASCAL VOC mean0average0precision0(map) 70% 60% 50% 40% 30% 20% 10% Before deep
More informationObject detection with CNNs
Object detection with CNNs 80% PASCAL VOC mean0average0precision0(map) 70% 60% 50% 40% 30% 20% 10% Before CNNs After CNNs 0% 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 year Region proposals
More informationR-FCN++: Towards Accurate Region-Based Fully Convolutional Networks for Object Detection
The Thirty-Second AAAI Conference on Artificial Intelligence (AAAI-18) R-FCN++: Towards Accurate Region-Based Fully Convolutional Networks for Object Detection Zeming Li, 1 Yilun Chen, 2 Gang Yu, 2 Yangdong
More informationObject Detection Based on Deep Learning
Object Detection Based on Deep Learning Yurii Pashchenko AI Ukraine 2016, Kharkiv, 2016 Image classification (mostly what you ve seen) http://tutorial.caffe.berkeleyvision.org/caffe-cvpr15-detection.pdf
More informationREGION AVERAGE POOLING FOR CONTEXT-AWARE OBJECT DETECTION
REGION AVERAGE POOLING FOR CONTEXT-AWARE OBJECT DETECTION Kingsley Kuan 1, Gaurav Manek 1, Jie Lin 1, Yuan Fang 1, Vijay Chandrasekhar 1,2 Institute for Infocomm Research, A*STAR, Singapore 1 Nanyang Technological
More informationLecture 5: Object Detection
Object Detection CSED703R: Deep Learning for Visual Recognition (2017F) Lecture 5: Object Detection Bohyung Han Computer Vision Lab. bhhan@postech.ac.kr 2 Traditional Object Detection Algorithms Region-based
More informationSpatial Localization and Detection. Lecture 8-1
Lecture 8: Spatial Localization and Detection Lecture 8-1 Administrative - Project Proposals were due on Saturday Homework 2 due Friday 2/5 Homework 1 grades out this week Midterm will be in-class on Wednesday
More informationMULTI-SCALE OBJECT DETECTION WITH FEATURE FUSION AND REGION OBJECTNESS NETWORK. Wenjie Guan, YueXian Zou*, Xiaoqun Zhou
MULTI-SCALE OBJECT DETECTION WITH FEATURE FUSION AND REGION OBJECTNESS NETWORK Wenjie Guan, YueXian Zou*, Xiaoqun Zhou ADSPLAB/Intelligent Lab, School of ECE, Peking University, Shenzhen,518055, China
More informationFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun Presented by Tushar Bansal Objective 1. Get bounding box for all objects
More informationContent-Based Image Recovery
Content-Based Image Recovery Hong-Yu Zhou and Jianxin Wu National Key Laboratory for Novel Software Technology Nanjing University, China zhouhy@lamda.nju.edu.cn wujx2001@nju.edu.cn Abstract. We propose
More informationIndustrial Technology Research Institute, Hsinchu, Taiwan, R.O.C ǂ
Stop Line Detection and Distance Measurement for Road Intersection based on Deep Learning Neural Network Guan-Ting Lin 1, Patrisia Sherryl Santoso *1, Che-Tsung Lin *ǂ, Chia-Chi Tsai and Jiun-In Guo National
More informationReal-time Object Detection CS 229 Course Project
Real-time Object Detection CS 229 Course Project Zibo Gong 1, Tianchang He 1, and Ziyi Yang 1 1 Department of Electrical Engineering, Stanford University December 17, 2016 Abstract Objection detection
More informationUnified, real-time object detection
Unified, real-time object detection Final Project Report, Group 02, 8 Nov 2016 Akshat Agarwal (13068), Siddharth Tanwar (13699) CS698N: Recent Advances in Computer Vision, Jul Nov 2016 Instructor: Gaurav
More informationEnd-to-End Airplane Detection Using Transfer Learning in Remote Sensing Images
remote sensing Article End-to-End Airplane Detection Using Transfer Learning in Remote Sensing Images Zhong Chen 1,2,3, Ting Zhang 1,2,3 and Chao Ouyang 1,2,3, * 1 School of Automation, Huazhong University
More informationarxiv: v1 [cs.cv] 5 Oct 2015
Efficient Object Detection for High Resolution Images Yongxi Lu 1 and Tara Javidi 1 arxiv:1510.01257v1 [cs.cv] 5 Oct 2015 Abstract Efficient generation of high-quality object proposals is an essential
More informationRegionlet Object Detector with Hand-crafted and CNN Feature
Regionlet Object Detector with Hand-crafted and CNN Feature Xiaoyu Wang Research Xiaoyu Wang Research Ming Yang Horizon Robotics Shenghuo Zhu Alibaba Group Yuanqing Lin Baidu Overview of this section Regionlet
More informationYiqi Yan. May 10, 2017
Yiqi Yan May 10, 2017 P a r t I F u n d a m e n t a l B a c k g r o u n d s Convolution Single Filter Multiple Filters 3 Convolution: case study, 2 filters 4 Convolution: receptive field receptive field
More informationComputer Vision Lecture 16
Computer Vision Lecture 16 Deep Learning Applications 11.01.2017 Bastian Leibe RWTH Aachen http://www.vision.rwth-aachen.de leibe@vision.rwth-aachen.de Announcements Seminar registration period starts
More informationFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks Shaoqing Ren Kaiming He Ross Girshick Jian Sun Present by: Yixin Yang Mingdong Wang 1 Object Detection 2 1 Applications Basic
More informationChannel Locality Block: A Variant of Squeeze-and-Excitation
Channel Locality Block: A Variant of Squeeze-and-Excitation 1 st Huayu Li Northern Arizona University Flagstaff, United State Northern Arizona University hl459@nau.edu arxiv:1901.01493v1 [cs.lg] 6 Jan
More informationMask R-CNN. presented by Jiageng Zhang, Jingyao Zhan, Yunhan Ma
Mask R-CNN presented by Jiageng Zhang, Jingyao Zhan, Yunhan Ma Mask R-CNN Background Related Work Architecture Experiment Mask R-CNN Background Related Work Architecture Experiment Background From left
More informationComputer Vision Lecture 16
Announcements Computer Vision Lecture 16 Deep Learning Applications 11.01.2017 Seminar registration period starts on Friday We will offer a lab course in the summer semester Deep Robot Learning Topic:
More informationarxiv: v1 [cs.cv] 15 Oct 2018
Instance Segmentation and Object Detection with Bounding Shape Masks Ha Young Kim 1,2,*, Ba Rom Kang 2 1 Department of Financial Engineering, Ajou University Worldcupro 206, Yeongtong-gu, Suwon, 16499,
More informationPedestrian Detection with Deep Convolutional Neural Network
Pedestrian Detection with Deep Convolutional Neural Network Xiaogang Chen, Pengxu Wei, Wei Ke, Qixiang Ye, Jianbin Jiao School of Electronic,Electrical and Communication Engineering, University of Chinese
More informationA FRAMEWORK OF EXTRACTING MULTI-SCALE FEATURES USING MULTIPLE CONVOLUTIONAL NEURAL NETWORKS. Kuan-Chuan Peng and Tsuhan Chen
A FRAMEWORK OF EXTRACTING MULTI-SCALE FEATURES USING MULTIPLE CONVOLUTIONAL NEURAL NETWORKS Kuan-Chuan Peng and Tsuhan Chen School of Electrical and Computer Engineering, Cornell University, Ithaca, NY
More informationDirect Multi-Scale Dual-Stream Network for Pedestrian Detection Sang-Il Jung and Ki-Sang Hong Image Information Processing Lab.
[ICIP 2017] Direct Multi-Scale Dual-Stream Network for Pedestrian Detection Sang-Il Jung and Ki-Sang Hong Image Information Processing Lab., POSTECH Pedestrian Detection Goal To draw bounding boxes that
More informationFeature-Fused SSD: Fast Detection for Small Objects
Feature-Fused SSD: Fast Detection for Small Objects Guimei Cao, Xuemei Xie, Wenzhe Yang, Quan Liao, Guangming Shi, Jinjian Wu School of Electronic Engineering, Xidian University, China xmxie@mail.xidian.edu.cn
More informationModern Convolutional Object Detectors
Modern Convolutional Object Detectors Faster R-CNN, R-FCN, SSD 29 September 2017 Presented by: Kevin Liang Papers Presented Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
More informationMimicking Very Efficient Network for Object Detection
Mimicking Very Efficient Network for Object Detection Quanquan Li 1, Shengying Jin 2, Junjie Yan 1 1 SenseTime 2 Beihang University liquanquan@sensetime.com, jsychffy@gmail.com, yanjunjie@outlook.com Abstract
More informationFaster R-CNN Features for Instance Search
Faster R-CNN Features for Instance Search Amaia Salvador, Xavier Giró-i-Nieto, Ferran Marqués Universitat Politècnica de Catalunya (UPC) Barcelona, Spain {amaia.salvador,xavier.giro}@upc.edu Shin ichi
More informationSupplementary material for Analyzing Filters Toward Efficient ConvNet
Supplementary material for Analyzing Filters Toward Efficient Net Takumi Kobayashi National Institute of Advanced Industrial Science and Technology, Japan takumi.kobayashi@aist.go.jp A. Orthonormal Steerable
More informationTRANSPARENT OBJECT DETECTION USING REGIONS WITH CONVOLUTIONAL NEURAL NETWORK
TRANSPARENT OBJECT DETECTION USING REGIONS WITH CONVOLUTIONAL NEURAL NETWORK 1 Po-Jen Lai ( 賴柏任 ), 2 Chiou-Shann Fuh ( 傅楸善 ) 1 Dept. of Electrical Engineering, National Taiwan University, Taiwan 2 Dept.
More informationObject detection using Region Proposals (RCNN) Ernest Cheung COMP Presentation
Object detection using Region Proposals (RCNN) Ernest Cheung COMP790-125 Presentation 1 2 Problem to solve Object detection Input: Image Output: Bounding box of the object 3 Object detection using CNN
More informationarxiv: v1 [cs.cv] 26 Jun 2017
Detecting Small Signs from Large Images arxiv:1706.08574v1 [cs.cv] 26 Jun 2017 Zibo Meng, Xiaochuan Fan, Xin Chen, Min Chen and Yan Tong Computer Science and Engineering University of South Carolina, Columbia,
More informationPT-NET: IMPROVE OBJECT AND FACE DETECTION VIA A PRE-TRAINED CNN MODEL
PT-NET: IMPROVE OBJECT AND FACE DETECTION VIA A PRE-TRAINED CNN MODEL Yingxin Lou 1, Guangtao Fu 2, Zhuqing Jiang 1, Aidong Men 1, and Yun Zhou 2 1 Beijing University of Posts and Telecommunications, Beijing,
More informationTowards Real-Time Automatic Number Plate. Detection: Dots in the Search Space
Towards Real-Time Automatic Number Plate Detection: Dots in the Search Space Chi Zhang Department of Computer Science and Technology, Zhejiang University wellyzhangc@zju.edu.cn Abstract Automatic Number
More informationEfficient Segmentation-Aided Text Detection For Intelligent Robots
Efficient Segmentation-Aided Text Detection For Intelligent Robots Junting Zhang, Yuewei Na, Siyang Li, C.-C. Jay Kuo University of Southern California Outline Problem Definition and Motivation Related
More informationYield Estimation using faster R-CNN
Yield Estimation using faster R-CNN 1 Vidhya Sagar, 2 Sailesh J.Jain and 2 Arjun P. 1 Assistant Professor, 2 UG Scholar, Department of Computer Engineering and Science SRM Institute of Science and Technology,Chennai,
More informationarxiv: v1 [cs.cv] 19 Feb 2019
Detector-in-Detector: Multi-Level Analysis for Human-Parts Xiaojie Li 1[0000 0001 6449 2727], Lu Yang 2[0000 0003 3857 3982], Qing Song 2[0000000346162200], and Fuqiang Zhou 1[0000 0001 9341 9342] arxiv:1902.07017v1
More informationObject Detection. CS698N Final Project Presentation AKSHAT AGARWAL SIDDHARTH TANWAR
Object Detection CS698N Final Project Presentation AKSHAT AGARWAL SIDDHARTH TANWAR Problem Description Arguably the most important part of perception Long term goals for object recognition: Generalization
More informationarxiv: v1 [cs.cv] 30 Apr 2018
An Anti-fraud System for Car Insurance Claim Based on Visual Evidence Pei Li Univeristy of Notre Dame BingYu Shen University of Notre dame Weishan Dong IBM Research China arxiv:184.1127v1 [cs.cv] 3 Apr
More informationSSD: Single Shot MultiBox Detector. Author: Wei Liu et al. Presenter: Siyu Jiang
SSD: Single Shot MultiBox Detector Author: Wei Liu et al. Presenter: Siyu Jiang Outline 1. Motivations 2. Contributions 3. Methodology 4. Experiments 5. Conclusions 6. Extensions Motivation Motivation
More informationObject Detection on Self-Driving Cars in China. Lingyun Li
Object Detection on Self-Driving Cars in China Lingyun Li Introduction Motivation: Perception is the key of self-driving cars Data set: 10000 images with annotation 2000 images without annotation (not
More informationReal-time object detection towards high power efficiency
Real-time object detection towards high power efficiency Jincheng Yu, Kaiyuan Guo, Yiming Hu, Xuefei Ning, Jiantao Qiu, Huizi Mao, Song Yao, Tianqi Tang, Boxun Li, Yu Wang, and Huazhong Yang Tsinghua University,
More informationTraffic Multiple Target Detection on YOLOv2
Traffic Multiple Target Detection on YOLOv2 Junhong Li, Huibin Ge, Ziyang Zhang, Weiqin Wang, Yi Yang Taiyuan University of Technology, Shanxi, 030600, China wangweiqin1609@link.tyut.edu.cn Abstract Background
More informationObject Detection. TA : Young-geun Kim. Biostatistics Lab., Seoul National University. March-June, 2018
Object Detection TA : Young-geun Kim Biostatistics Lab., Seoul National University March-June, 2018 Seoul National University Deep Learning March-June, 2018 1 / 57 Index 1 Introduction 2 R-CNN 3 YOLO 4
More informationPixel Offset Regression (POR) for Single-shot Instance Segmentation
Pixel Offset Regression (POR) for Single-shot Instance Segmentation Yuezun Li 1, Xiao Bian 2, Ming-ching Chang 1, Longyin Wen 2 and Siwei Lyu 1 1 University at Albany, State University of New York, NY,
More informationFine-tuning Pre-trained Large Scaled ImageNet model on smaller dataset for Detection task
Fine-tuning Pre-trained Large Scaled ImageNet model on smaller dataset for Detection task Kyunghee Kim Stanford University 353 Serra Mall Stanford, CA 94305 kyunghee.kim@stanford.edu Abstract We use a
More informationExtend the shallow part of Single Shot MultiBox Detector via Convolutional Neural Network
Extend the shallow part of Single Shot MultiBox Detector via Convolutional Neural Network Liwen Zheng, Canmiao Fu, Yong Zhao * School of Electronic and Computer Engineering, Shenzhen Graduate School of
More informationMask R-CNN. Kaiming He, Georgia, Gkioxari, Piotr Dollar, Ross Girshick Presenters: Xiaokang Wang, Mengyao Shi Feb. 13, 2018
Mask R-CNN Kaiming He, Georgia, Gkioxari, Piotr Dollar, Ross Girshick Presenters: Xiaokang Wang, Mengyao Shi Feb. 13, 2018 1 Common computer vision tasks Image Classification: one label is generated for
More informationarxiv: v3 [cs.cv] 18 Oct 2017
SSH: Single Stage Headless Face Detector Mahyar Najibi* Pouya Samangouei* Rama Chellappa University of Maryland arxiv:78.3979v3 [cs.cv] 8 Oct 27 najibi@cs.umd.edu Larry S. Davis {pouya,rama,lsd}@umiacs.umd.edu
More informationSmart Parking System using Deep Learning. Sheece Gardezi Supervised By: Anoop Cherian Peter Strazdins
Smart Parking System using Deep Learning Sheece Gardezi Supervised By: Anoop Cherian Peter Strazdins Content Labeling tool Neural Networks Visual Road Map Labeling tool Data set Vgg16 Resnet50 Inception_v3
More information[Supplementary Material] Improving Occlusion and Hard Negative Handling for Single-Stage Pedestrian Detectors
[Supplementary Material] Improving Occlusion and Hard Negative Handling for Single-Stage Pedestrian Detectors Junhyug Noh Soochan Lee Beomsu Kim Gunhee Kim Department of Computer Science and Engineering
More informationarxiv: v2 [cs.cv] 23 Nov 2017
Light-Head R-CNN: In Defense of Two-Stage Object Detector Zeming Li 1, Chao Peng 2, Gang Yu 2, Xiangyu Zhang 2, Yangdong Deng 1, Jian Sun 2 1 School of Software, Tsinghua University, {lizm15@mails.tsinghua.edu.cn,
More informationFCHD: A fast and accurate head detector
JOURNAL OF L A TEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 1 FCHD: A fast and accurate head detector Aditya Vora, Johnson Controls Inc. arxiv:1809.08766v2 [cs.cv] 26 Sep 2018 Abstract In this paper, we
More informationA Novel Representation and Pipeline for Object Detection
A Novel Representation and Pipeline for Object Detection Vishakh Hegde Stanford University vishakh@stanford.edu Manik Dhar Stanford University dmanik@stanford.edu Abstract Object detection is an important
More informationComputer Vision Lecture 16
Computer Vision Lecture 16 Deep Learning for Object Categorization 14.01.2016 Bastian Leibe RWTH Aachen http://www.vision.rwth-aachen.de leibe@vision.rwth-aachen.de Announcements Seminar registration period
More informationFinal Report: Smart Trash Net: Waste Localization and Classification
Final Report: Smart Trash Net: Waste Localization and Classification Oluwasanya Awe oawe@stanford.edu Robel Mengistu robel@stanford.edu December 15, 2017 Vikram Sreedhar vsreed@stanford.edu Abstract Given
More informationProceedings of the International MultiConference of Engineers and Computer Scientists 2018 Vol I IMECS 2018, March 14-16, 2018, Hong Kong
, March 14-16, 2018, Hong Kong , March 14-16, 2018, Hong Kong , March 14-16, 2018, Hong Kong , March 14-16, 2018, Hong Kong TABLE I CLASSIFICATION ACCURACY OF DIFFERENT PRE-TRAINED MODELS ON THE TEST DATA
More informationAdaptive Object Detection Using Adjacency and Zoom Prediction
Adaptive Object Detection Using Adjacency and Zoom Prediction Yongxi Lu University of California, San Diego yol7@ucsd.edu Tara Javidi University of California, San Diego tjavidi@ucsd.edu Svetlana Lazebnik
More informationKaggle Data Science Bowl 2017 Technical Report
Kaggle Data Science Bowl 2017 Technical Report qfpxfd Team May 11, 2017 1 Team Members Table 1: Team members Name E-Mail University Jia Ding dingjia@pku.edu.cn Peking University, Beijing, China Aoxue Li
More informationDeepVoting: A Robust and Explainable Deep Network for Semantic Part Detection under Partial Occlusion
CBMM Memo No. 083 Jun 19, 018 DeepVoting: A Robust and Explainable Deep Network for Semantic Part Detection under Partial Occlusion Zhishuai Zhang 1, Cihang Xie 1, Jianyu Wang, Lingxi Xie 1, Alan L. Yuille
More informationYOLO9000: Better, Faster, Stronger
YOLO9000: Better, Faster, Stronger Date: January 24, 2018 Prepared by Haris Khan (University of Toronto) Haris Khan CSC2548: Machine Learning in Computer Vision 1 Overview 1. Motivation for one-shot object
More informationRich feature hierarchies for accurate object detection and semantic segmentation
Rich feature hierarchies for accurate object detection and semantic segmentation BY; ROSS GIRSHICK, JEFF DONAHUE, TREVOR DARRELL AND JITENDRA MALIK PRESENTER; MUHAMMAD OSAMA Object detection vs. classification
More informationPedestrian Detection based on Deep Fusion Network using Feature Correlation
Pedestrian Detection based on Deep Fusion Network using Feature Correlation Yongwoo Lee, Toan Duc Bui and Jitae Shin School of Electronic and Electrical Engineering, Sungkyunkwan University, Suwon, South
More informationDeepProposals: Hunting Objects and Actions by Cascading Deep Convolutional Layers
DOI.7/s263-7-6-x DeepProposals: Hunting Objects and Actions by Cascading Deep Convolutional Layers Amir Ghodrati Ali Diba 2 Marco Pedersoli 3 Tinne Tuytelaars 2 Luc Van Gool 2 Received: 3 May 26 / Accepted:
More informationEFFECTIVE OBJECT DETECTION FROM TRAFFIC CAMERA VIDEOS. Honghui Shi, Zhichao Liu*, Yuchen Fan, Xinchao Wang, Thomas Huang
EFFECTIVE OBJECT DETECTION FROM TRAFFIC CAMERA VIDEOS Honghui Shi, Zhichao Liu*, Yuchen Fan, Xinchao Wang, Thomas Huang Image Formation and Processing (IFP) Group, University of Illinois at Urbana-Champaign
More informationarxiv: v1 [cs.cv] 31 Mar 2016
Object Boundary Guided Semantic Segmentation Qin Huang, Chunyang Xia, Wenchao Zheng, Yuhang Song, Hao Xu and C.-C. Jay Kuo arxiv:1603.09742v1 [cs.cv] 31 Mar 2016 University of Southern California Abstract.
More informationDeepBox: Learning Objectness with Convolutional Networks
DeepBox: Learning Objectness with Convolutional Networks Weicheng Kuo Bharath Hariharan Jitendra Malik University of California, Berkeley {wckuo, bharath2, malik}@eecs.berkeley.edu Abstract Existing object
More informationG-CNN: an Iterative Grid Based Object Detector
G-CNN: an Iterative Grid Based Object Detector Mahyar Najibi 1, Mohammad Rastegari 1,2, Larry S. Davis 1 1 University of Maryland, College Park 2 Allen Institute for Artificial Intelligence najibi@cs.umd.edu
More informationarxiv: v4 [cs.cv] 6 Jul 2016
Object Boundary Guided Semantic Segmentation Qin Huang, Chunyang Xia, Wenchao Zheng, Yuhang Song, Hao Xu, C.-C. Jay Kuo (qinhuang@usc.edu) arxiv:1603.09742v4 [cs.cv] 6 Jul 2016 Abstract. Semantic segmentation
More informationarxiv: v1 [cs.cv] 4 Jun 2015
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks arxiv:1506.01497v1 [cs.cv] 4 Jun 2015 Shaoqing Ren Kaiming He Ross Girshick Jian Sun Microsoft Research {v-shren, kahe, rbg,
More informationImproving Small Object Detection
Improving Small Object Detection Harish Krishna, C.V. Jawahar CVIT, KCIS International Institute of Information Technology Hyderabad, India Abstract While the problem of detecting generic objects in natural
More informationDefect Detection from UAV Images based on Region-Based CNNs
Defect Detection from UAV Images based on Region-Based CNNs Meng Lan, Yipeng Zhang, Lefei Zhang, Bo Du School of Computer Science, Wuhan University, Wuhan, China {menglan, yp91, zhanglefei, remoteking}@whu.edu.cn
More informationDeep Tracking: Biologically Inspired Tracking with Deep Convolutional Networks
Deep Tracking: Biologically Inspired Tracking with Deep Convolutional Networks Si Chen The George Washington University sichen@gwmail.gwu.edu Meera Hahn Emory University mhahn7@emory.edu Mentor: Afshin
More informationR-FCN: OBJECT DETECTION VIA REGION-BASED FULLY CONVOLUTIONAL NETWORKS
R-FCN: OBJECT DETECTION VIA REGION-BASED FULLY CONVOLUTIONAL NETWORKS JIFENG DAI YI LI KAIMING HE JIAN SUN MICROSOFT RESEARCH TSINGHUA UNIVERSITY MICROSOFT RESEARCH MICROSOFT RESEARCH SPEED/ACCURACY TRADE-OFFS
More informationDeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution and Fully Connected CRFs
DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution and Fully Connected CRFs Zhipeng Yan, Moyuan Huang, Hao Jiang 5/1/2017 1 Outline Background semantic segmentation Objective,
More informationHand Detection For Grab-and-Go Groceries
Hand Detection For Grab-and-Go Groceries Xianlei Qiu Stanford University xianlei@stanford.edu Shuying Zhang Stanford University shuyingz@stanford.edu Abstract Hands detection system is a very critical
More informationOBJECT DETECTION HYUNG IL KOO
OBJECT DETECTION HYUNG IL KOO INTRODUCTION Computer Vision Tasks Classification + Localization Classification: C-classes Input: image Output: class label Evaluation metric: accuracy Localization Input:
More informationObject Detection. Part1. Presenter: Dae-Yong
Object Part1 Presenter: Dae-Yong Contents 1. What is an Object? 2. Traditional Object Detector 3. Deep Learning-based Object Detector What is an Object? Subset of Object Recognition What is an Object?
More informationClassification of objects from Video Data (Group 30)
Classification of objects from Video Data (Group 30) Sheallika Singh 12665 Vibhuti Mahajan 12792 Aahitagni Mukherjee 12001 M Arvind 12385 1 Motivation Video surveillance has been employed for a long time
More informationarxiv: v2 [cs.cv] 8 Apr 2018
Single-Shot Object Detection with Enriched Semantics Zhishuai Zhang 1 Siyuan Qiao 1 Cihang Xie 1 Wei Shen 1,2 Bo Wang 3 Alan L. Yuille 1 Johns Hopkins University 1 Shanghai University 2 Hikvision Research
More informationConvolutional Neural Networks. Computer Vision Jia-Bin Huang, Virginia Tech
Convolutional Neural Networks Computer Vision Jia-Bin Huang, Virginia Tech Today s class Overview Convolutional Neural Network (CNN) Training CNN Understanding and Visualizing CNN Image Categorization:
More informationHIERARCHICAL JOINT-GUIDED NETWORKS FOR SEMANTIC IMAGE SEGMENTATION
HIERARCHICAL JOINT-GUIDED NETWORKS FOR SEMANTIC IMAGE SEGMENTATION Chien-Yao Wang, Jyun-Hong Li, Seksan Mathulaprangsan, Chin-Chin Chiang, and Jia-Ching Wang Department of Computer Science and Information
More informationarxiv: v2 [cs.cv] 10 Apr 2017
Fully Convolutional Instance-aware Semantic Segmentation Yi Li 1,2 Haozhi Qi 2 Jifeng Dai 2 Xiangyang Ji 1 Yichen Wei 2 1 Tsinghua University 2 Microsoft Research Asia {liyi14,xyji}@tsinghua.edu.cn, {v-haoq,jifdai,yichenw}@microsoft.com
More informationStructured Prediction using Convolutional Neural Networks
Overview Structured Prediction using Convolutional Neural Networks Bohyung Han bhhan@postech.ac.kr Computer Vision Lab. Convolutional Neural Networks (CNNs) Structured predictions for low level computer
More informationAn Analysis of Scale Invariance in Object Detection SNIP
An Analysis of Scale Invariance in Object Detection SNIP Bharat Singh Larry S. Davis University of Maryland, College Park {bharat,lsd}@cs.umd.edu Abstract An analysis of different techniques for recognizing
More informationarxiv: v1 [cs.cv] 26 May 2017
arxiv:1705.09587v1 [cs.cv] 26 May 2017 J. JEONG, H. PARK AND N. KWAK: UNDER REVIEW IN BMVC 2017 1 Enhancement of SSD by concatenating feature maps for object detection Jisoo Jeong soo3553@snu.ac.kr Hyojin
More informationTexture Complexity based Redundant Regions Ranking for Object Proposal
26 IEEE Conference on Computer Vision and Pattern Recognition Workshops Texture Complexity based Redundant Regions Ranking for Object Proposal Wei Ke,2, Tianliang Zhang., Jie Chen 2, Fang Wan, Qixiang
More informationA MultiPath Network for Object Detection
ZAGORUYKO et al.: A MULTIPATH NETWORK FOR OBJECT DETECTION 1 A MultiPath Network for Object Detection Sergey Zagoruyko, Adam Lerer, Tsung-Yi Lin, Pedro O. Pinheiro, Sam Gross, Soumith Chintala, Piotr Dollár
More informationarxiv: v2 [cs.cv] 18 Jul 2017
PHAM, ITO, KOZAKAYA: BISEG 1 arxiv:1706.02135v2 [cs.cv] 18 Jul 2017 BiSeg: Simultaneous Instance Segmentation and Semantic Segmentation with Fully Convolutional Networks Viet-Quoc Pham quocviet.pham@toshiba.co.jp
More informationHENet: A Highly Efficient Convolutional Neural. Networks Optimized for Accuracy, Speed and Storage
HENet: A Highly Efficient Convolutional Neural Networks Optimized for Accuracy, Speed and Storage Qiuyu Zhu Shanghai University zhuqiuyu@staff.shu.edu.cn Ruixin Zhang Shanghai University chriszhang96@shu.edu.cn
More informationObject Detection for Crime Scene Evidence Analysis using Deep Learning
Object Detection for Crime Scene Evidence Analysis using Deep Learning Surajit Saikia 1,2, E. Fidalgo 1,2, Enrique Alegre 1,2 and 2,3 Laura Fernández-Robles 1 Department of Electrical, Systems and Automation,
More informationarxiv: v2 [cs.cv] 19 Apr 2018
arxiv:1804.06215v2 [cs.cv] 19 Apr 2018 DetNet: A Backbone network for Object Detection Zeming Li 1, Chao Peng 2, Gang Yu 2, Xiangyu Zhang 2, Yangdong Deng 1, Jian Sun 2 1 School of Software, Tsinghua University,
More informationarxiv: v3 [cs.cv] 2 Jun 2017
Incorporating the Knowledge of Dermatologists to Convolutional Neural Networks for the Diagnosis of Skin Lesions arxiv:1703.01976v3 [cs.cv] 2 Jun 2017 Iván González-Díaz Department of Signal Theory and
More informationObject Detection with YOLO on Artwork Dataset
Object Detection with YOLO on Artwork Dataset Yihui He Computer Science Department, Xi an Jiaotong University heyihui@stu.xjtu.edu.cn Abstract Person: 0.64 Horse: 0.28 I design a small object detection
More informationLearning to Segment Object Candidates
Learning to Segment Object Candidates Pedro O. Pinheiro Ronan Collobert Piotr Dollár pedro@opinheiro.com locronan@fb.com pdollar@fb.com Facebook AI Research Abstract Recent object detection systems rely
More information2017 2nd International Conference on Software, Multimedia and Communication Engineering (SMCE 2017) ISBN:
2017 2nd International Conference on Software, Multimedia and Communication Engineering (SMCE 2017) ISBN: 978-1-60595-458-5 Aircraft Detection in Remote Sensing Images via CNN Multi-scale Feature Representation
More informationarxiv: v1 [cs.cv] 22 Mar 2018
Single-Shot Bidirectional Pyramid Networks for High-Quality Object Detection Xiongwei Wu, Daoxin Zhang, Jianke Zhu, Steven C.H. Hoi School of Information Systems, Singapore Management University, Singapore
More informationObject Detection and Its Implementation on Android Devices
Object Detection and Its Implementation on Android Devices Zhongjie Li Stanford University 450 Serra Mall, Stanford, CA 94305 jay2015@stanford.edu Rao Zhang Stanford University 450 Serra Mall, Stanford,
More information