Instance Segmentation On Lvis V1 0 Val

评估指标

mask AP

mask APr

评测结果

各个模型在此基准测试上的表现结果

模型名称	mask AP	mask APr	Paper Title	Repository
ViTDet-H	48.1	36.9	Exploring Plain Vision Transformer Backbones for Object Detection
R101-FPN-MaskRCNN-EQLv2	27.5	-	Equalization Loss v2: A New Gradient Balance Approach for Long-tailed Object Detection
SE-R50-FPN-MaskRCNN-APA	29.1	21.6	Adaptive Parametric Activation	-
FRACAL-SwinS	33.6	27.8	Fractal Calibration for long-tailed object detection
DiffusionInst-ResNet50	22.3	-	DiffusionInst: Diffusion Model for Instance Segmentation
SE-R101-FPN-MaskRCNN-APA	30.7	23.6	Adaptive Parametric Activation	-
DiffusionInst-SwinB	36	-	DiffusionInst: Diffusion Model for Instance Segmentation
ViTDet-L	46.0	34.3	Exploring Plain Vision Transformer Backbones for Object Detection
DiffusionInst-SwinL	38.6	-	DiffusionInst: Diffusion Model for Instance Segmentation
GLEE-Pro	49.9	-	General Object Foundation Model for Images and Videos at Scale
R50-FPN-MaskRCNN-GOL	27.7	-	Long-tailed Instance Segmentation using Gumbel Optimized Loss
R101-MaskRCNN-LOCE	28	-	Exploring Classification Equilibrium in Long-Tailed Object Detection	-
DiverGen (Swin-L)	45.5	45.8	DiverGen: Improving Instance Segmentation by Learning Wider Data Distribution with More Diverse Generative Data
R50-FPN-MaskRCNN-EQLv2	23.7	-	Equalization Loss v2: A New Gradient Balance Approach for Long-tailed Object Detection
CenterNet2 (Swin-L w/ X-Paste + Copy-Paste)	45.4	43.8	X-Paste: Revisiting Scalable Copy-Paste for Instance Segmentation using CLIP and StableDiffusion
Eff-B7 NAS-FPN (1280, Copy-Paste pre-training))	38.1	-	Simple Copy-Paste is a Strong Data Augmentation Method for Instance Segmentation
Co-DETR (single-scale)	60.7	-	DETRs with Collaborative Hybrid Assignments Training
R50-MaskRCNN-LOCE	26.6	-	Exploring Classification Equilibrium in Long-Tailed Object Detection	-
HTC-ealrystop-ema	50.4	-	LVIS Challenge Track Technical Report 1st Place Solution: Distribution Balanced and Boundary Refinement for Large Vocabulary Instance Segmentation	-
Cascade-R101-EQLv2	28.8	-	Equalization Loss v2: A New Gradient Balance Approach for Long-tailed Object Detection

0 of 24 row(s) selected.