List your company

Scene Parsing on Cityscapes (test)

80.4mIoUSFNet (ResNet-18) with Mapillary pretraining
57.41663.38369.3575.317Feb 24, 2020

Evaluation results

MethodLinks
SFNet (ResNet-18) with Mapillary pretrainingBackbone=ResNet-18, InputSize=1024 x 2048, Pre-training=Mapillary dataset, Note=FPS represents without/with TensorRT acceleration2020.0280.4——
SFNet (ResNet-18)Backbone=ResNet-18, InputSize=1024 x 2048, Note=FPS represents without/with TensorRT acceleration2020.0278.9——
SFNet (DF2)Backbone=DF2, InputSize=1024 x 2048, Note=FPS represents without/with TensorRT acceleration2020.0277.8——
SwiftNet-ensInputSize=1024 x 20482020.0276.518.4—
SwiftNetInputSize=1024 x 20482020.0275.539.9—
DF2-SegInputSize=1024 x 20482020.0274.855—
BiSeNetBackbone=ResNet-18, InputSize=768 x 15362020.0274.643—
SFNet (DF1)Backbone=DF1, InputSize=1024 x 2048, Note=FPS represents without/with TensorRT acceleration2020.0274.5——
DF1-SegInputSize=1024 x 20482020.027380—
DFANetInputSize=1024 x 10242020.0271.3100—
CellNetInputSize=768 x 15362020.0270.5108—
ERFNetInputSize=512 x 10242020.0269.741.9—
ICNetInputSize=1024 x 20482020.0269.534—
BiSeNetBackbone=Xception-39, InputSize=768 x 15362020.0268.472—
ESPNetv2InputSize=512 x 10242020.0262.180—
ESPNetInputSize=512 x 10242020.0260.3132—
ENetInputSize=640 x 3602020.0258.360—