Rethinking Atrous Convolution for Semantic Image Segmentation

In this work, we revisit atrous convolution, a powerful tool to explicitly adjust filter's field-of-view as well as control the resolution of feature responses computed by Deep Convolutional Neural Networks, in the application of semantic image segmentation. To handle the problem of segmenting objects at multiple scales, we design modules which employ atrous convolution in cascade or in parallel to capture multi-scale context by adopting multiple atrous rates... (read more)

PDF Abstract

Results from the Paper


Ranked #6 on Semantic Segmentation on PASCAL VOC 2012 val (using extra training data)

     Get a GitHub badge
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK USES EXTRA
TRAINING DATA
RESULT BENCHMARK
Semantic Segmentation Cityscapes test DeepLabv3 (ResNet-101, coarse) Mean IoU (class) 81.3% # 25
Semantic Segmentation Cityscapes val DeepLabv3 (Dilated-ResNet-101) mIoU 78.5% # 15
Semantic Segmentation PASCAL VOC 2012 test DeepLabv3-JFT Mean IoU 86.9% # 6
Semantic Segmentation PASCAL VOC 2012 val DeepLabv3-JFT mIoU 82.70% # 6

Methods used in the Paper


METHOD TYPE
Spatial Pyramid Pooling
Pooling Operations
Average Pooling
Pooling Operations
ASPP
Semantic Segmentation Modules
SGD with Momentum
Stochastic Optimization
Weight Decay
Regularization
Random Horizontal Flip
Image Data Augmentation
Random Scaling
Image Data Augmentation
Polynomial Rate Decay
Learning Rate Schedules
DeepLabv3
Semantic Segmentation Models
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
ResNet
Convolutional Neural Networks
Convolution
Convolutions