HGFormer Model Zoo and Baselines

It's common to initialize from backbone models pre-trained on ImageNet classification tasks.

To prepare the backbones pre-trained on ImageNet classification, please following tools/README.md

All models available for download through this document are licensed under the Creative Commons Attribution-NonCommercial 4.0 International License.

Cityscapes -> ACDC

Method	Backbone	Fog	Night	Rain	Snow	All	Download
Mask2former	Swin-Tiny	54.06	38.11	59.54	55.76	53.65	model
HGFormer	Swin-Tiny	59.82	41.88	60.92	60.82	56.95	model

Method	Backbone	Average	Motion	Defoc	Glass	Gauss	Gauss	Impul	Shot	Speck	Bright	Contr	Satur	JPEG	Snow	Spatt	Fog	Frost	Download
Mask2former	Swin-Tiny	41.68	51.61	51.52	39.69	46.71	6.89	7.68	12.75	44.10	72.71	58.60	69.14	22.86	26.10	58.35	67.12	31.11	model
HGFormer	Swin-Tiny	43.81	52.51	53.03	39.02	47.93	16.45	16.03	20.55	48.44	74.51	57.14	70.53	27.32	25.66	59.19	66.49	26.11	model

Method	Backbone	Mapillary	BDD	GTA	Synthia	Average	Download
Mask2former	Swin-Tiny	65.28	49.87	51.38	34.76	50.32	model
HGFormer	Swin-Tiny	67.22	52.69	51.94	32.98	51.21	model

Method	Backbone	GTA	Synthia	Cityscapes	BDD	Average	Download
Mask2former	Swin-Tiny	57.81	40.14	68.23	59.05	56.31	model
HGFormer	Swin-Tiny	60.79	39.15	69.28	62.22	57.86	model

The numbers differ slightly from the results reported in the paper because we presented an average of three runs in the paper.