Instance Segmentation On Coco

Metrics

AP50

AP75

APL

APM

APS

mask AP

Results

Performance results of various models on this benchmark

Model Name	AP50	AP75	APL	APM	APS	mask AP	Paper Title	Repository
CenterMask + VoVNetV2-99 (single-scale)	62.3	44.1	57.0	42.8	20.1	40.6	CenterMask : Real-Time Anchor-Free Instance Segmentation	-
MasK DINO (SwinL, multi-scale)	-	-	-	-	-	54.7	Mask DINO: Towards A Unified Transformer-based Framework for Object Detection and Segmentation	-
ISDA (ours)	62	41.1	-	41.2	17	38.7	ISDA: Position-Aware Instance Segmentation with Deformable Attention	-
EmbedMask(R-101-FPN)	59.1	40.3	-	40.4	17.9	37.7	EmbedMask: Embedding Coupling for One-stage Instance Segmentation	-
PANet	-	-	-	-	-	42.0	Path Aggregation Network for Instance Segmentation	-
GLEE-Lite	-	-	-	-	-	48.3	General Object Foundation Model for Images and Videos at Scale	-
VirTex Mask R-CNN (ResNet-50-FPN)	58.4	39.7	-	-	-	36.9	VirTex: Learning Visual Representations from Textual Annotations	-
iBOT (ViT-B/16)	-	-	-	-	-	44.2	iBOT: Image BERT Pre-Training with Online Tokenizer	-
DiffusionInst-ResNet101	-	-	-	-	-	41.5	DiffusionInst: Diffusion Model for Instance Segmentation	-
Cascade Mask R-CNN (ResNeXt152, CBNet)	-	-	-	-	-	43.3	CBNet: A Novel Composite Backbone Network Architecture for Object Detection	-
ViT-Adapter-L (HTC++, BEiTv2 pretrain, multi-scale)	-	-	-	-	-	53.0	Vision Transformer Adapter for Dense Predictions	-
PolarMask (ResNet-101-FPN)	51.9%	31%	42.8%	32.4%	13.4%	30.4%	PolarMask: Single Shot Instance Segmentation with Polar Representation	-
Co-DETR	80.2	63.4	72.0	60.1	41.6	57.1	DETRs with Collaborative Hybrid Assignments Training	-
gSwin-S	-	-	-	-	-	45.03	gSwin: Gated MLP Vision Model with Hierarchical Structure of Shifted Window	-
MogaNet-B (Cascade Mask R-CNN)	-	-	-	-	-	46	MogaNet: Multi-order Gated Aggregation Network	-
DetectoRS (ResNeXt-101-32x4d, multi-scale)	71.1	51.6	59.6	49.5	30.3	47.1	DetectoRS: Detecting Objects with Recursive Feature Pyramid and Switchable Atrous Convolution	-
VoVNetV1-57	-	-	-	-	-	40.8%	An Energy and GPU-Computation Efficient Backbone Network for Real-Time Object Detection	-
GCNet (ResNeXt-101 + DCN + cascade + GC r16)	-	-	-	-	-	41.5%	GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond	-
SOLQ (ResNet50, single scale)	-	-	-	-	-	39.7	SOLQ: Segmenting Objects by Learning Queries	-
dBOT ViT-B (CLIP)	-	-	-	-	-	46.2	Exploring Target Representations for Masked Autoencoders	-

0 of 112 row(s) selected.