ppt sppnet_ilsvrc2014.pdf 金字塔pooling spp！Microsoft

文件名称: ppt sppnet_ilsvrc2014.pdf

所属分类: 深度学习

开发工具:

文件大小: 344kb

下载次数: 0

上传时间: 2019-08-16

提供者: hzh***

下载 (344kb)

不能下载？报告错误

详细说明：ppt sppnet_ilsvrc2014.pdf 金字塔pooling spp！Microsoft Research Overview ·SPP-net a new network structure Classification improves all CNNs Detection 20-60x faster than r-CNN, as accurate ECQyJ4 Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition K. He, X Zhang, S. Ren, J. Sun ECCV 2014 Microsoft Research Spatial Pyramid Matching SPM: very successful in traditional computer vision Grauman Darrell, ICCV 2005The Pyramid Match Kernel: Discriminative Classification with Sets of Image Features [Lazebnik et al, CVPR 2006]"Beyond Bags of Features: Spatial Pyramid Matching for Recognizing Natural Scene Categories → → prediction dense sift encoded SPⅣ SVM VQ, SC, FV CNN conv layers simply pooling? fc layers counterparts ECQyJ4 Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition K. He, X Zhang, S. Ren, J. Sun ECCV 2014 Microsoft Research SPP-net: spm in cnn traditional CNN fixed size cony fc SPP-net spatial pyramid any poolIng Fix bin numbers DO NOT fix bin size ECQyJ4 Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition K. He, X Zhang, S. Ren, J. Sun ECCV 2014 Microsoft Research SPP-net c layers variable input size scale multi-size training concatenate multi-scale testing full-image view multi-level pooling robust to deformation operates on feature maps pooling in regions spatial pyramid pooling layer conv feature maps conv layers Input image △B4 Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition K. He, X Zhang, S. Ren, J. Sun ECCV 2014 Microsoft LSVRC top-5 val (10-view) Research no-SPP baselines 15.00 14.76 14.14 1392 14.00 multi-level pooling 13.50 13.54 1352 13.64 13.33 1280 12.50 12.00 12.33 multi-size training 11.97 11.50 All CNNS 11.12 11.00 improved! 1095 10.00 4 architectures ZF-5 Conynet*5 Overfeat-5 Overfeat-7 ECQyJ4 Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition K. He, X Zhang, S. Ren, J. Sun ECCV 2014 Microsoft Research LSVRC 2014 CLS Results team top-5 test 666 7-cony SPP-net, 10-view 10.95% GoogLeNet Oxford vgg 7.32 ours 8.06 7-cony SPP-net, 96-view+2-full 9.08% Howard 8.11 Deeper vision 9.50 multiple SPP-nets 8.06% NUS-BST 9.79 TTIC ECP 10.22 “ shallow 7-conv, 1 Titan gpu, 3 weeks but potential SPP can improve deeper nets: >1% gain post-competition △B4 Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition K. He, X Zhang, S. Ren, J. Sun ECCV 2014 Microsoft Research Detection: SPP on Regions fc layers SPP cony feature maps region conv layers input image △B4 Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition K. He, X Zhang, S. Ren, J. Sun ECCV 2014 Microsoft Research cnN VS SPP image regions vs. feature map regions feature feature feature feature feature feature feature net net net Image Image R-CNN SPP-net 2000 nets on image regions 1 net on full image △B4 Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition K. He, X Zhang, S. Ren, J. Sun ECCV 2014

(系统自动生成,下载前可以参看下载内容)