Instance Segmentation On Coco
評価指標
AP50
AP75
APL
APM
APS
mask AP
評価結果
このベンチマークにおける各モデルのパフォーマンス結果
比較表
| モデル名 | AP50 | AP75 | APL | APM | APS | mask AP |
|---|---|---|---|---|---|---|
| centermask-real-time-anchor-free-instance-1 | 62.3 | 44.1 | 57.0 | 42.8 | 20.1 | 40.6 |
| mask-dino-towards-a-unified-transformer-based-1 | - | - | - | - | - | 54.7 |
| isda-position-aware-instance-segmentation | 62 | 41.1 | - | 41.2 | 17 | 38.7 |
| embedmask-embedding-coupling-for-one-stage | 59.1 | 40.3 | - | 40.4 | 17.9 | 37.7 |
| path-aggregation-network-for-instance | - | - | - | - | - | 42.0 |
| general-object-foundation-model-for-images | - | - | - | - | - | 48.3 |
| virtex-learning-visual-representations-from | 58.4 | 39.7 | - | - | - | 36.9 |
| ibot-image-bert-pre-training-with-online | - | - | - | - | - | 44.2 |
| diffusioninst-diffusion-model-for-instance | - | - | - | - | - | 41.5 |
| cbnet-a-novel-composite-backbone-network | - | - | - | - | - | 43.3 |
| vision-transformer-adapter-for-dense | - | - | - | - | - | 53.0 |
| polarmask-single-shot-instance-segmentation | 51.9% | 31% | 42.8% | 32.4% | 13.4% | 30.4% |
| detrs-with-collaborative-hybrid-assignments | 80.2 | 63.4 | 72.0 | 60.1 | 41.6 | 57.1 |
| gswin-gated-mlp-vision-model-with | - | - | - | - | - | 45.03 |
| efficient-multi-order-gated-aggregation | - | - | - | - | - | 46 |
| detectors-detecting-objects-with-recursive-1 | 71.1 | 51.6 | 59.6 | 49.5 | 30.3 | 47.1 |
| an-energy-and-gpu-computation-efficient | - | - | - | - | - | 40.8% |
| gcnet-non-local-networks-meet-squeeze | - | - | - | - | - | 41.5% |
| solq-segmenting-objects-by-learning-queries | - | - | - | - | - | 39.7 |
| exploring-target-representations-for-masked | - | - | - | - | - | 46.2 |
| hybrid-task-cascade-for-instance-segmentation | - | - | - | - | - | 41.2% |
| mask-r-cnn | 60.0 | 39.4 | 53.5 | 39.9 | 16.9 | 37.1 |
| rdsnet-a-new-deep-architecture-for-reciprocal | 57.9% | 39.0% | 51.6% | 39.5% | 16.4% | 36.4% |
| ibot-image-bert-pre-training-with-online | - | - | - | - | - | 42.6 |
| centermask-real-time-anchor-free-instance-1 | 60.8 | - | - | 41.7 | 19.4 | - |
| focal-self-attention-for-local-global | 75.4 | 56.5 | 64.2 | - | 35.6 | 51.3 |
| architecture-agnostic-masked-image-modeling | - | - | - | - | - | 43.5 |
| centermask-real-time-anchor-free-instance-1 | 66.2 | 47.4 | - | - | 27.2 | - |
| embedmask-embedding-coupling-for-one-stage | 59.1% | 40.3% | 53% | 40.4% | 17.9% | 37.7% |
| k-net-towards-unified-image-segmentation | 62.8 | - | 58.8 | 42.7 | 18.7 | 40.1% |
| a-tri-layer-plugin-to-improve-occluded | - | - | - | - | - | 45.9 |
| mask-dino-towards-a-unified-transformer-based-1 | - | - | - | - | - | 52.8 |
| polarmask-enhanced-polar-representation-for | 64.1 | 40.0 | 52.0 | 40.2 | 22.2 | 38.7 |
| efficient-multi-order-gated-aggregation | - | - | - | - | - | 35.8 |
| centermask-real-time-anchor-free-instance-1 | - | - | - | - | - | 38.3 |
| eva-exploring-the-limits-of-masked-visual | 80.0 | - | 72.4 | 58.0 | 36.3 | 55.5 |
| istr-end-to-end-instance-segmentation-with | - | - | 52.3 | 41.9 | 22.8 | 39.9% |
| solo-segmenting-objects-by-locations | 62.7 | 43.3 | 58.9 | 43.3 | 17.6 | 40.4 |
| swin-transformer-hierarchical-vision | - | - | - | - | - | 51.1 |
| exploring-target-representations-for-masked | - | - | - | - | - | 48.8 |
| k-net-towards-unified-image-segmentation | 63.3 | - | 59 | 43.3 | 18.8 | 40.6% |
| resnest-split-attention-networks | - | - | - | - | - | 43% |
| istr-end-to-end-instance-segmentation-with | - | - | - | - | - | 49.7 |
| isda-position-aware-instance-segmentation | - | - | 55.7 | - | - | - |
| fully-convolutional-instance-aware-semantic | 54.5% | - | - | - | - | 33.6% |
| efficient-multi-order-gated-aggregation | - | - | - | - | - | 43.2 |
| diffusioninst-diffusion-model-for-instance | - | - | - | - | - | 37.1 |
| image-as-a-foreign-language-beit-pretraining | - | - | - | - | - | 54.8 |
| solq-segmenting-objects-by-learning-queries | - | - | - | - | - | 40.9 |
| fully-convolutional-instance-aware-semantic | 49.5% | - | 50.0% | 31.3% | 7.1% | 29.2% |
| general-object-foundation-model-for-images | - | - | - | - | - | 53.3 |
| instaboost-boosting-instance-segmentation-via | 61.4% | 42.9% | 52.1% | 42.5% | 21.2% | 39.5% |
| centermask-real-time-anchor-free-instance-1 | - | - | 54.3 | 44.4 | 24.4 | 41.8 |
| efficient-multi-order-gated-aggregation | - | - | - | - | - | 46.1 |
| d2det-towards-high-quality-object-detection | 61.5 | 43.7 | 54.0 | 43.0 | 21.7 | 40.2 |
| architecture-agnostic-masked-image-modeling | - | - | - | - | - | 34.9 |
| spinenet-learning-scale-permuted-backbone-for | - | - | - | - | - | 46.1 |
| efficient-multi-order-gated-aggregation | - | - | - | - | - | 42.2 |
| general-object-foundation-model-for-images | - | - | - | - | - | 54.5 |
| commonality-parsing-network-across-shape-and | 60.8 | 42.2 | 50.1 | 41.8 | 22.2 | 39.2 |
| istr-end-to-end-instance-segmentation-with | - | - | 50.6 | 40.4 | 22.1 | 38.6% |
| diffusioninst-diffusion-model-for-instance | - | - | - | - | - | 47.6 |
| mask-scoring-r-cnn | - | - | - | - | - | 39.6% |
| solov2-dynamic-faster-and-stronger | 63.2 | 45.1 | 61.6 | 45.0 | 18.0 | 41.7 |
| cbnetv2-a-composite-backbone-network | - | - | - | - | - | 51.6 |
| centermask-real-time-anchor-free-instance-1 | 61.2 | 42.9 | - | - | 19.7 | 39.6 |
| torchdistill-a-modular-configuration-driven | - | - | - | - | - | 33.6 |
| exploring-target-representations-for-masked | - | - | - | - | - | 48.3 |
| e2ec-an-end-to-end-contour-based-method-for | 52.9 | 35.9 | - | - | - | 33.8 |
| masklab-instance-segmentation-by-refining | - | - | - | - | - | 38.1% |
| sipmask-spatial-information-preservation-for | 60.2 | 40.8 | 54.3 | 40.8 | 17.8 | 38.1 |
| queryinst-parallelly-supervised-mask-query | 74.2 | 53.8 | 63.2 | 51.8 | 31.5 | 49.1 |
| swin-transformer-v2-scaling-up-capacity-and | - | - | - | - | - | 54.4 |
| deep-occlusion-aware-instance-segmentation | - | - | - | - | - | 41.7 |
| universal-instance-perception-as-object | 76.2 | 56.7 | 67.5 | 55.9 | 33.3 | 51.8 |
| hybrid-task-cascade-for-instance-segmentation | - | - | - | - | - | 41.2 |
| exploring-target-representations-for-masked | - | - | - | - | - | 46.3 |
| global-context-networks | 68.9 | 49.6 | - | - | - | 45.4 |
| contrastive-learning-rivals-masked-image | - | - | - | - | - | 55.4 |
| efficient-multi-order-gated-aggregation | - | - | - | - | - | 45.1 |
| solo-segmenting-objects-by-locations | 62.7% | 43.3% | 58.9% | 43.3% | 17.6% | 40.4% |
| an-energy-and-gpu-computation-efficient | - | - | - | - | - | 39.7% |
| deep-occlusion-aware-instance-segmentation | 61.2 | 42.7 | 51.0 | 42.3 | 22.3 | 39.6 |
| vision-transformer-adapter-for-dense | - | - | - | - | - | 52.5 |
| cbnetv2-a-composite-backbone-network | - | - | - | - | - | 52.3 |
| simple-copy-paste-is-a-strong-data | - | - | - | - | - | 46.9 |
| simple-copy-paste-is-a-strong-data | - | - | - | - | - | 49.1 |
| vision-transformer-adapter-for-dense | - | - | - | - | - | 54.5 |
| efficient-multi-order-gated-aggregation | - | - | - | - | - | 44.1 |
| diffusioninst-diffusion-model-for-instance | - | - | - | - | - | 48.3 |
| detectors-detecting-objects-with-recursive-1 | 72.0 | 53.3 | 61.5 | 50.9 | 31.6 | 48.5 |
| resnest-split-attention-networks | 70.2 | 51.5 | 60.6 | 49.6 | 30.0 | - |
| swin-transformer-hierarchical-vision | - | - | - | - | - | 50.2 |
| efficient-multi-order-gated-aggregation | - | - | - | - | - | 37.6 |
| yolact-real-time-instance-segmentation | - | - | - | - | - | 29.8% |
| efficient-multi-order-gated-aggregation | - | - | - | - | - | 48.8 |
| a-multipath-network-for-object-detection | - | - | - | - | - | 25.0% |
| tensormask-a-foundation-for-dense-object | - | - | - | - | - | 37.3% |
| gswin-gated-mlp-vision-model-with | - | - | - | - | - | 42.87 |
| mask-transfiner-for-high-quality-instance | - | - | - | - | - | 42.2 |
| gswin-gated-mlp-vision-model-with | - | - | - | - | - | 44.16 |
| deep-occlusion-aware-instance-segmentation | 61.5 | 43.1 | 51.1 | 42.4 | 22.7 | 39.8 |
| internimage-exploring-large-scale-vision | 80.8 | 62.2 | 70.3 | 58.9 | 41.0 | - |
| instance-aware-semantic-segmentation-via | 44.3% | - | - | - | - | - |
| blendmask-top-down-meets-bottom-up-for | 63.1 | 44.6 | 54.5 | 44.1 | 22.7 | 41.3 |
| mask-frozen-detr-high-quality-instance | 79.3 | 61.4 | 70.4 | 58.4 | 37.8 | 55.3 |
| end-to-end-semi-supervised-object-detection | - | - | - | - | - | 53.0 |
| cbnetv2-a-composite-backbone-network | 80.3 | 62.1 | 70.9 | 59.3 | 39.7 | 56.1 |
| solq-segmenting-objects-by-learning-queries | - | - | - | - | - | 46.7 |
| masked-attention-mask-transformer-for | 74.9 | 54.9 | 71.2 | 53.8 | 29.1 | 50.5 |
| polarmask-single-shot-instance-segmentation | 55.4% | 33.8% | 46.3% | 35.1% | 15.5% | 32.9% |
| efficient-multi-order-gated-aggregation | - | - | - | - | - | 39.1 |