Pyramid Scene Parsing Network

Scene parsing is challenging for unrestricted open vocabulary and diverse scenes. In this paper, we exploit the capability of global context information by different-region-based context aggregation through our pyramid pooling module together with the proposed pyramid scene parsing network (PSPNet)... (read more)

PDF Abstract CVPR 2017 PDF CVPR 2017 Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK USES EXTRA
TRAINING DATA
RESULT BENCHMARK
Semantic Segmentation ADE20K PSPNet Validation mIoU 44.94 # 11
Test Score 55.38 # 7
Semantic Segmentation ADE20K val PSPNet (ResNet-101) mIoU 43.29 # 23
Semantic Segmentation ADE20K val PSPNet (ResNet-152) mIoU 43.51 # 22
Lesion Segmentation Anatomical Tracings of Lesions After Stroke (ATLAS) PSPNet Dice 0.3571 # 7
IoU 0.2540 # 6
Precision 0.4769 # 7
Recall 0.3335 # 7
Semantic Segmentation CamVid PSPNet Mean IoU 69.1% # 3
Real-Time Semantic Segmentation CamVid PSPNet mIoU 69.1% # 9
Time (ms) 185 # 8
Frame (fps) 5.4 # 4
Semantic Segmentation Cityscapes test PSPNet (ResNet-101) Mean IoU (class) 78.4% # 37
Semantic Segmentation Cityscapes val PSPNet (Dilated-ResNet-101) mIoU 79.7% # 12
Video Semantic Segmentation Cityscapes val PSPNet-101 [20] mIoU 79.7 # 3
Video Semantic Segmentation Cityscapes val PSPNet-50 [20] mIoU 78.1 # 4
Real-Time Semantic Segmentation NYU Depth v2 PSPNet101 mIoU 43.2 # 4
Speed(ms/f) 72 # 8
Real-Time Semantic Segmentation NYU Depth v2 PSPNet50 mIoU 41.8 # 6
Speed(ms/f) 47 # 7
Real-Time Semantic Segmentation NYU Depth v2 PSPNet18 mIoU 35.9 # 9
Speed(ms/f) 19 # 2
Semantic Segmentation PASCAL Context PSPNet (ResNet-101) mIoU 47.8 # 28
Semantic Segmentation PASCAL VOC 2012 test PSPNet Mean IoU 85.4% # 12
Semantic Segmentation PASCAL VOC 2012 test PSPNet (ResNet-101) Mean IoU 82.6% # 23

Results from Other Papers


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK USES EXTRA
TRAINING DATA
SOURCE PAPER COMPARE
Semantic Segmentation Cityscapes test PSPNet (ResNet-101) Mean IoU (class) 81.2% # 27

Methods used in the Paper


METHOD TYPE
Average Pooling
Pooling Operations
Auxiliary Classifier
Miscellaneous Components
Pyramid Pooling Module
Semantic Segmentation Modules
FCN
Semantic Segmentation Models
Random Gaussian Blur
Image Data Augmentation
RandomRotate
Image Data Augmentation
Random Horizontal Flip
Image Data Augmentation
Weight Decay
Regularization
SGD with Momentum
Stochastic Optimization
Polynomial Rate Decay
Learning Rate Schedules
PSPNet
Semantic Segmentation Models
Dilated Convolution
Convolutions
Residual Connection
Skip Connections
ReLU
Activation Functions
1x1 Convolution
Convolutions
Batch Normalization
Normalization
Bottleneck Residual Block
Skip Connection Blocks
Global Average Pooling
Pooling Operations
Residual Block
Skip Connection Blocks
Kaiming Initialization
Initialization
Max Pooling
Pooling Operations
Convolution
Convolutions
ResNet
Convolutional Neural Networks