Semantic Segmentation On Pascal Voc 2012

평가 지표

Mean IoU

평가 결과

이 벤치마크에서 각 모델의 성능 결과

모델 이름	Mean IoU	Paper Title	Repository
FCN (VGG-16)	62.2%	Fully Convolutional Networks for Semantic Segmentation
Dilated FCN-2s VGG19	69%	Efficient Yet Deep Convolutional Neural Networks for Semantic Segmentation
Light-Weight-RefineNet-50	81.1%	Light-Weight RefineNet for Real-Time Semantic Segmentation
Light-Weight-RefineNet-101	82.0%	Light-Weight RefineNet for Real-Time Semantic Segmentation
DeepLabv3+ (Xception-JFT)	89.0%	Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation
Smooth Network with Channel Attention Block	86.2%	Learning a Discriminative Feature Network for Semantic Segmentation
SANet (pretraining on COCO dataset)	86.1%	Squeeze-and-Attention Networks for Semantic Segmentation
ESPNet	63.01%	ESPNet: Efficient Spatial Pyramid of Dilated Convolutions for Semantic Segmentation
TuSimple	83.1%	Understanding Convolution for Semantic Segmentation
PSPNet	85.4%	Pyramid Scene Parsing Network
ParseNet	69.8%	ParseNet: Looking Wider to See Better
G2	56.7%	Exploiting saliency for object segmentation from image level labels	-
CentraleSupelec Deep G-CRF	80.2%	Fast, Exact and Multi-Scale Inference for Semantic Image Segmentation with Deep Gaussian CRFs
WASPnet-CRF (ours)	79.6%	Waterfall Atrous Spatial Pooling Architecture for Efficient Semantic Segmentation
EncNet (ResNet-101)	82.9%	Context Encoding for Semantic Segmentation
SSDD	65.5	Self-Supervised Difference Detection for Weakly-Supervised Semantic Segmentation
Dilated Convolutions	67.6%	Multi-Scale Context Aggregation by Dilated Convolutions
CASIA_IVA_SDN	86.6%	Stacked Deconvolutional Network for Semantic Segmentation	-
ESPNetv2	68.0%	ESPNetv2: A Light-weight, Power Efficient, and General Purpose Convolutional Neural Network
Deeplab-v2 with Lovasz-Softmax loss	79.00%	The Lovász-Softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks

0 of 51 row(s) selected.