Semantic Segmentation On Bdd100K Val

Metrics

mIoU

Results

Performance results of various models on this benchmark

Model Name	mIoU	Paper Title
Deeplabv3+	63.6	Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation
DSNet-Base	64.6	DSNet: A Novel Way to Use Atrous Convolutions in Semantic Segmentation
DF1-Seg	42.5(82.3fps)	Partial Order Pruning: for Best Speed/Accuracy Trade-off in Neural Architecture Search
SERNet-Former_v2	67.42	SERNet-Former: Semantic Segmentation by Efficient Residual Network with Attention-Boosting Gates and Attention-Fusion Networks
Bi-Align	53.4(42.1fps)	Fast and Accurate Scene Parsing via Bi-direction Alignment Networks
EMANet	61.4	Expectation-Maximization Attention Networks for Semantic Segmentation
BiSeNet-V1(ResNet-18)	53.8(45.1fps)	BiSeNet: Bilateral Segmentation Network for Real-time Semantic Segmentation
OCRNet	60.1	Segmentation Transformer: Object-Contextual Representations for Semantic Segmentation
PSPNet	62.3	Pyramid Scene Parsing Network
SFNet-Lite(STDC2)	60.6(194.5FPS 4090)	SFNet: Faster, Accurate, and Domain Agnostic Semantic Segmentation via Semantic Flow
ICNet	52.4(39.5fps)	ICNet for Real-Time Semantic Segmentation on High-Resolution Images
DF2-Seg	47.8(53.4fps)	Partial Order Pruning: for Best Speed/Accuracy Trade-off in Neural Architecture Search
NiseNet	53.52	What's There in the Dark
SFNet(ResNet-18)	60.6(132.5FPS 4090)	Semantic Flow for Fast and Accurate Scene Parsing
VLTSeg	72.5	Strong but simple: A Baseline for Domain Generalized Dense Perception by CLIP-based Transfer Learning
SFNet(DF1)	55.4(70.3fps)	Semantic Flow for Fast and Accurate Scene Parsing
STDC1	52.1(45.8FPS)	Rethinking BiSeNet For Real-time Semantic Segmentation
SFNet-Lite(ResNet-18)	60.6(161.3FPS 4090)	SFNet: Faster, Accurate, and Domain Agnostic Semantic Segmentation via Semantic Flow
DSNet-head64	62.6(172.2FPS 4090)	DSNet: A Novel Way to Use Atrous Convolutions in Semantic Segmentation
STDC2	53.8(33.0FPS)	Rethinking BiSeNet For Real-time Semantic Segmentation

0 of 24 row(s) selected.