Panoptic Segmentation On Coco Minival

Results

Performance results of various models on this benchmark

Comparison Table

Model Name	AP	PQ
focal-modulation-networks	48.4	57.9
a-simple-framework-for-open-vocabulary	53.2	59.5
visual-attention-network	-	58.2
fully-convolutional-networks-for-panoptic	-	44.3
per-pixel-classification-is-not-all-you-need	-	52.7
k-means-mask-transformer	-	58.0
end-to-end-object-detection-with-transformers	33	45.1
axial-deeplab-stand-alone-axial-attention-for	-	43.9
end-to-end-object-detection-with-transformers	39.7	44.1
k-means-mask-transformer	-	58.1
mask-dino-towards-a-unified-transformer-based-1	50.9	59.4
oneformer-one-transformer-to-rule-universal	52.0	60.0
axial-deeplab-stand-alone-axial-attention-for	-	-
k-means-mask-transformer	-	57.9
dilated-neighborhood-attention-transformer	49.2	58.5
oneformer-one-transformer-to-rule-universal	49.0	57.9
hyperseg-towards-universal-visual	-	61.2
axial-deeplab-stand-alone-axial-attention-for	-	43.4
oneformer-one-transformer-to-rule-universal	49.2	58.0
vision-transformer-adapter-for-dense	48.9	58.4
fully-convolutional-networks-for-panoptic	-	-
panoptic-segformer	-	55.8
max-deeplab-end-to-end-panoptic-segmentation	-	51.1
masked-attention-mask-transformer-for	48.6	57.8
hierarchical-open-vocabulary-universal-image-1	-	58.1
cmt-deeplab-clustering-mask-transformers-for-1	-	55.3
resnest-split-attention-networks	-	47.9
panoptic-segformer	-	50.6