List your company
ResearchFormerly WizWand

Semantic Segmentation on ADE20K (fine-tuning) (mIoU)

54.4mIoUSoftmax
9.773621.359332.94544.5307Mar 17, 2026
Updated 7mo ago

Evaluation results

MethodLinks
SoftmaxBackbone=SigLIP-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=310.60, Peak Mem. (GB)=1.3181, Throughput (imgs/s)=36.52, Segmentation Head=Mask2former2026.0354.4
ViT-AdaLA (Ours)Backbone=SigLIP-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=262.19↓15.6%, Peak Mem. (GB)=1.2163↓7.7%, Throughput (imgs/s)=41.56↑16.1%, Segmentation Head=Mask2former2026.0353.55
ViT-AdaLA (Stage 2)Backbone=SigLIP-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=262.19↓15.6%, Peak Mem. (GB)=1.2163↓7.7%, Throughput (imgs/s)=41.56↑16.1%, Segmentation Head=Mask2former2026.0351.39
SoftmaxBackbone=IN1K ViT-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=310.60, Peak Mem. (GB)=1.3181, Throughput (imgs/s)=36.52, Segmentation Head=Mask2former2026.0350.83
ViT-AdaLA (Ours)Backbone=IN1K ViT-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=262.19↓15.6%, Peak Mem. (GB)=1.2163↓7.7%, Throughput (imgs/s)=41.56↑16.1%, Segmentation Head=Mask2former2026.0349.75
ViT-AdaLA (Stage 2)Backbone=IN1K ViT-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=262.19↓15.6%, Peak Mem. (GB)=1.2163↓7.7%, Throughput (imgs/s)=41.56↑16.1%, Segmentation Head=Mask2former2026.0346.92
MonarchBackbone=IN1K ViT-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=269.44↓13.3%, Peak Mem. (GB)=1.2304↓6.7%, Throughput (imgs/s)=18.31↓48.9%, Segmentation Head=Mask2former2026.0345.05
NystromformerBackbone=IN1K ViT-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=265.20↓14.6%, Peak Mem. (GB)=1.2163↓7.7%, Throughput (imgs/s)=18.29↓50.8%, Segmentation Head=Mask2former2026.0341.67
MonarchBackbone=SigLIP-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=269.44↓13.3%, Peak Mem. (GB)=1.2304↓6.7%, Throughput (imgs/s)=18.31↓48.9%, Segmentation Head=Mask2former2026.0338.87
NystromformerBackbone=SigLIP-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=265.20↓14.6%, Peak Mem. (GB)=1.2163↓7.7%, Throughput (imgs/s)=18.29↓50.8%, Segmentation Head=Mask2former2026.0332.54
PerformerBackbone=IN1K ViT-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=265.41↓14.6%, Peak Mem. (GB)=1.2164↓7.7%, Throughput (imgs/s)=36.48↑0.1%, Segmentation Head=Mask2former2026.0331.01
CosformerBackbone=IN1K ViT-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=265.41↓14.6%, Peak Mem. (GB)=1.2226↓7.2%, Throughput (imgs/s)=39.55↑8.2%, Segmentation Head=Mask2former2026.0329.65
HedgehogBackbone=IN1K ViT-L, Res.=512^2, Params (M)=305.77, FLOPS (G)=265.41↓14.5%, Peak Mem. (GB)=1.2221↓7.3%, Throughput (imgs/s)=37.44↑2.5%, Segmentation Head=Mask2former2026.0325.15
PerformerBackbone=SigLIP-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=265.41↓14.6%, Peak Mem. (GB)=1.2164↓7.7%, Throughput (imgs/s)=36.48↑0.1%, Segmentation Head=Mask2former2026.0323.97
CosformerBackbone=SigLIP-L, Res.=512^2, Params (M)=304.20, FLOPS (G)=265.41↓14.6%, Peak Mem. (GB)=1.2226↓7.2%, Throughput (imgs/s)=39.55↑8.2%, Segmentation Head=Mask2former2026.0323.08
HedgehogBackbone=SigLIP-L, Res.=512^2, Params (M)=305.77, FLOPS (G)=265.41↓14.5%, Peak Mem. (GB)=1.2221↓7.3%, Throughput (imgs/s)=37.44↑2.5%, Segmentation Head=Mask2former2026.0322.13
LoLCATSBackbone=IN1K ViT-L, Res.=512^2, Params (M)=305.77, FLOPS (G)=265.41↓14.5%, Peak Mem. (GB)=1.2221↓7.3%, Throughput (imgs/s)=37.44↑2.5%, Segmentation Head=Mask2former2026.0320.17
LoLCATSBackbone=SigLIP-L, Res.=512^2, Params (M)=305.77, FLOPS (G)=265.41↓14.5%, Peak Mem. (GB)=1.2221↓7.3%, Throughput (imgs/s)=37.44↑2.5%, Segmentation Head=Mask2former2026.0320.11
LinformerBackbone=SigLIP-L, Res.=512^2, Params (M)=305.77, FLOPS (G)=262.19↓15.6%, Peak Mem. (GB)=1.2221↓7.3%, Throughput (imgs/s)=45.41↑25.3%, Segmentation Head=Mask2former2026.0312.37
LinformerBackbone=IN1K ViT-L, Res.=512^2, Params (M)=305.77, FLOPS (G)=262.19↓15.6%, Peak Mem. (GB)=1.2221↓7.3%, Throughput (imgs/s)=45.41↑25.3%, Segmentation Head=Mask2former2026.0311.49