Object Detection On Coco
評価指標
AP50
AP75
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 |
|---|---|---|
| dynamic-head-unifying-object-detection-heads | 64.5 | 50.7 |
| yolov7-trainable-bag-of-freebies-sets-new | - | - |
| efficientdet-scalable-and-efficient-object | 71.6 | 56.9 |
| generalized-focal-loss-v2-learning-reliable | 67.6 | 53.5 |
| revisiting-unreasonable-effectiveness-of-data | 58 | 40.1 |
| grounded-language-image-pre-training | 79.5 | 67.7 |
| group-detr-v2-strong-object-detector-with-1 | 81.8 | 71.1 |
| mask-r-cnn | 62.3 | 43.4 |
| istr-end-to-end-instance-segmentation-with | - | - |
| corner-proposal-network-for-anchor-free-two | 67.3 | 53.7 |
| m2det-a-single-shot-object-detector-based-on | 64.6 | 49.3 |
| yolov7-trainable-bag-of-freebies-sets-new | - | - |
| architecture-agnostic-masked-image-modeling | - | - |
| an-analysis-of-scale-invariance-in-object-1 | 65.5 | 48.4 |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| centermask-real-time-anchor-free-instance-1 | 61.6 | 46.9 |
| attention-guided-context-feature-pyramid | 70.4 | 57 |
| mnasfpn-learning-latency-aware-pyramid | - | - |
| localization-uncertainty-estimation-for | - | - |
| speedaccuracy-trade-offs-for-modern | - | - |
| generalized-focal-loss-v2-learning-reliable | 66.5 | 52.8 |
| mnasfpn-learning-latency-aware-pyramid | - | - |
| single-shot-refinement-neural-network-for | 57.5 | 39.5 |
| yolox-exceeding-yolo-series-in-2021 | - | - |
| deep-residual-learning-for-image-recognition | - | - |
| cbnetv2-a-composite-backbone-network | - | - |
| dynamic-head-unifying-object-detection-heads | 78.5 | 66.6 |
| path-aggregation-network-for-instance | 67.2 | 51.8 |
| fast-r-cnn | - | - |
| reducing-label-noise-in-anchor-free-object | 64.8 | 51.6 |
| deformable-kernels-adapting-effective | - | - |
| ibot-image-bert-pre-training-with-online | - | - |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| swin-transformer-hierarchical-vision | - | - |
| edgenext-efficiently-amalgamated-cnn | - | - |
| reppoints-point-set-representation-for-object | 67.4 | 50.9 |
| reppoints-point-set-representation-for-object | 65.0 | 46.3 |
| general-object-foundation-model-for-images | - | - |
| centermask-real-time-anchor-free-instance-1 | 63.1 | 48.6 |
| nas-fcos-fast-neural-architecture-search-for | - | - |
| towards-all-in-one-pre-training-via | - | - |
| you-only-look-one-level-feature | 62.9 | 47.5 |
| architecture-agnostic-masked-image-modeling | - | - |
| bottom-up-object-detection-by-grouping | 55.5 | 43.2 |
| exploring-target-representations-for-masked | - | - |
| fcos-fully-convolutional-one-stage-object | 62.2 | 46.1 |
| location-sensitive-visual-recognition-with | 71.1 | 59.2 |
| cross-iteration-batch-normalization | 60.5 | 44.1 |
| feature-selective-anchor-free-module-for | 65.2 | 48.6 |
| centernet-object-detection-with-keypoint | 64.5 | 50.7 |
| a-ranking-based-balanced-loss-function | 65.0 | 47.5 |
| learning-data-augmentation-strategies-for | - | - |
| torchdistill-a-modular-configuration-driven | - | - |
| lip-local-importance-based-pooling | 65.7 | 48.1 |
| usb-universal-scale-object-detection | 67.5 | 53.0 |
| image-as-a-foreign-language-beit-pretraining | - | - |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| cascade-r-cnn-delving-into-high-quality | 61.1 | 41.9 |
| chainercv-a-library-for-deep-learning-in | - | - |
| hybrid-task-cascade-for-instance-segmentation | 63.9 | 44.7 |
| probabilistic-anchor-assignment-with-iou | 71.6 | 59.1 |
| feature-pyramid-networks-for-object-detection | - | - |
| vision-transformer-with-deformable-attention | 69.6 | 51.2 |
| deformable-convnets-v2-more-deformable-better | 67.9 | 50.8 |
| 190807919 | 64.0 | 50.3 |
| feature-intertwiner-for-object-detection-1 | 67.5 | 51.1 |
| compact-global-descriptor-for-neural-networks | - | - |
| simple-copy-paste-is-a-strong-data | - | - |
| solq-segmenting-objects-by-learning-queries | 74.6 | 60.5 |
| inside-outside-net-detecting-objects-in | 55.7 | 34.6 |
| pp-yoloe-an-evolved-version-of-yolo | 68.9 | 55.6 |
| saccadenet-a-fast-and-accurate-object | 55.6 | 41.4 |
| m2det-a-single-shot-object-detector-based-on | 59.7 | 45 |
| spinenet-learning-scale-permuted-backbone-for | 68.4 | 52.5 |
| vision-transformer-adapter-for-dense | - | - |
| focal-loss-for-dense-object-detection | 59.1 | 42.3 |
| swin-transformer-v2-scaling-up-capacity-and | - | - |
| cascade-rpn-delving-into-high-quality-region | 58.9 | 44.5 |
| soft-anchor-point-object-detection | 67.4 | 51.1 |
| usb-universal-scale-object-detection | 71.6 | 59.9 |
| mask-r-cnn | 60.3 | 41.7 |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| attention-guided-context-feature-pyramid | 64.4 | 49 |
| relationnet-bridging-visual-representations | - | - |
| multiple-anchor-learning-for-visual-object | - | - |
| compact-global-descriptor-for-neural-networks | - | - |
| dino-detr-with-improved-denoising-anchor-1 | - | - |
| scale-aware-trident-networks-for-object | 63.6 | 46.5 |
| mobilenets-efficient-convolutional-neural | - | - |
| detrs-with-collaborative-hybrid-assignments | - | - |
| general-object-foundation-model-for-images | - | - |
| global-context-networks | 70.9 | 56.9 |
| focal-self-attention-for-local-global | - | - |
| you-only-learn-one-representation-unified | 73.3 | 60.6 |
| matrix-nets-a-new-deep-architecture-for | 66.2 | 52.3 |
| learning-spatial-fusion-for-single-shot | 64.1 | 49.2 |
| centernet-for-object-detection | 73.7 | 62.4 |
| cascade-r-cnn-delving-into-high-quality | 59 | 39.2 |
| 190807919 | 62.5 | 48.6 |
| understanding-gaussian-attention-bias-of | - | - |
| exploring-target-representations-for-masked | - | - |
| generalized-focal-loss-v2-learning-reliable | 69 | 55.3 |
| end-to-end-semi-supervised-object-detection | - | - |
| 190807919 | - | 46.5 |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| grit-a-generative-region-to-text-transformer | - | - |
| grid-r-cnn | 63.0 | 46.6 |
| internimage-exploring-large-scale-vision | - | - |
| fcos-fully-convolutional-one-stage-object | 62.8 | 46.6 |
| single-shot-refinement-neural-network-for | 54.5 | 35.5 |
| yolox-exceeding-yolo-series-in-2021 | - | - |
| region-proposal-by-guided-anchoring | 59.2 | 43.5 |
| d2det-towards-high-quality-object-detection | 69.4 | 54.9 |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| reppoints-point-set-representation-for-object | 62.9 | 44.3 |
| 190408900 | - | - |
| centermask-real-time-anchor-free-instance-1 | 64.5 | - |
| detrs-with-collaborative-hybrid-assignments | - | - |
| tood-task-aligned-one-stage-object-detection | 60.3 | 46.4 |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| instaboost-boosting-instance-segmentation-via | 64.2 | 50 |
| single-shot-refinement-neural-network-for | 62.9 | 45.7 |
| dynamic-r-cnn-towards-high-quality-object | 68.3 | 55.6 |
| mocae-mixture-of-calibrated-experts | - | - |
| objects-as-points | - | - |
| gcnet-non-local-networks-meet-squeeze | 67.6 | 52.7 |
| detr-does-not-need-multi-scale-or-locality | 82.1 | 70.7 |
| yolox-exceeding-yolo-series-in-2021 | - | - |
| eva-exploring-the-limits-of-masked-visual | 81.9 | 71.7 |
| dynamic-head-unifying-object-detection-heads | 60.7 | 46.8 |
| 190408900 | - | - |
| generalized-focal-loss-v2-learning-reliable | 64.3 | 50.5 |
| dctd-deep-conditional-target-densities-for | 58.5 | 41.8 |
| fcos-fully-convolutional-one-stage-object | 64.1 | 48.4 |
| dynamic-head-unifying-object-detection-heads | 65.7 | 51.9 |
| 190409925 | - | - |
| feature-selective-anchor-free-module-for | 61.5 | 44 |
| torchdistill-a-modular-configuration-driven | - | - |
| glipv2-unifying-localization-and-vision | - | - |
| mnasfpn-learning-latency-aware-pyramid | - | - |
| cascade-r-cnn-delving-into-high-quality | 59.9 | 44 |
| general-object-foundation-model-for-images | - | - |
| foveabox-beyond-anchor-based-object-detector | - | - |
| espnetv2-a-light-weight-power-efficient-and | - | - |
| reppoints-v2-verification-meets-regression | 70.1 | 57.5 |
| dynamic-head-unifying-object-detection-heads | 77.1 | 64.5 |
| yolov7-trainable-bag-of-freebies-sets-new | - | - |
| m2det-a-single-shot-object-detector-based-on | 64.4 | 48 |
| 190807919 | 59.3 | - |
| a-strong-and-reproducible-object-detector | 81.7 | 71.5 |
| mnasfpn-learning-latency-aware-pyramid | - | - |
| nas-fcos-fast-neural-architecture-search-for | - | - |
| scaled-yolov4-scaling-cross-stage-partial | 73.2 | 61.2 |
| freeanchor-learning-to-match-anchors-for | 64.3 | 48.4 |
| retinamask-learning-to-predict-masks-improves | 62.5 | 46.0 |
| cbnetv2-a-composite-backbone-network | - | - |
| generalized-focal-loss-v2-learning-reliable | 62.3 | 48.5 |
| a-multipath-network-for-object-detection | - | - |
| scaled-yolov4-scaling-cross-stage-partial | 70.3 | 58 |
| deformable-detr-deformable-transformers-for-1 | 71.9 | 58.1 |
| detectors-detecting-objects-with-recursive-1 | 71.6 | 58.5 |
| foveabox-beyond-anchor-based-object-detector | - | - |
| generalized-focal-loss-learning-qualified-and | 67.4 | 52.6 |
| pp-yoloe-an-evolved-version-of-yolo | 69.9 | 56.5 |
| virtex-learning-visual-representations-from | 61.7 | 44.8 |
| solq-segmenting-objects-by-learning-queries | - | - |
| yolox-exceeding-yolo-series-in-2021 | - | - |
| scaled-yolov4-scaling-cross-stage-partial | 64.1 | 49.5 |
| scaled-yolov4-scaling-cross-stage-partial | 72.3 | 59.5 |
| vision-transformer-adapter-for-dense | - | - |
| cornernet-detecting-objects-as-paired | 57.8 | 45.3 |
| libra-r-cnn-towards-balanced-learning-for | 64 | 47 |
| a-ranking-based-balanced-loss-function | 68.4 | 51.1 |
| houghnet-integrating-near-and-long-range | 65.1 | 50.7 |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| scaled-yolov4-scaling-cross-stage-partial | 72.6 | 60.2 |
| revisiting-the-sibling-head-in-object | 69.6 | 54.4 |
| centermask-real-time-anchor-free-instance-1 | 68.3 | 53.2 |
| yolov7-trainable-bag-of-freebies-sets-new | - | - |
| focal-loss-for-dense-object-detection | 61.1 | 44.1 |
| focal-modulation-networks | - | - |
| segmentation-is-all-you-need | - | - |
| a-ranking-based-balanced-loss-function | 70.3 | 53.9 |
| 190807919 | 63.6 | 46.4 |
| spinenet-learning-scale-permuted-backbone-for | 60.5 | 44.6 |
| relation-detr-exploring-explicit-position | 80.8 | 69.1 |
| yolov4-optimal-speed-and-accuracy-of-object | 65.7 | 47.3 |
| detectors-detecting-objects-with-recursive-1 | 74.2 | 61.1 |
| resnest-split-attention-networks | 72.0 | 58.0 |
| spinenet-learning-scale-permuted-backbone-for | 66.3 | 50.6 |
| contrastive-learning-rivals-masked-image | - | - |
| sniper-efficient-multi-scale-training | 65.0 | 48.6 |
| boosting-r-cnn-reweighting-r-cnn-samples-by | - | - |
| reversible-column-networks | - | - |
| multiple-anchor-learning-for-visual-object | - | - |
| beyond-skip-connections-top-down-modulation | - | - |
| spinenet-learning-scale-permuted-backbone-for | 63.8 | 47.6 |
| scale-equalizing-pyramid-convolution-for | 69.8 | 54.3 |
| foveabox-beyond-anchor-based-object-detector | 61.9 | 45.2 |
| ssd-single-shot-multibox-detector | 48.5 | 30.3 |
| bridging-the-gap-between-anchor-based-and | 68.9 | 56.3 |
| rdsnet-a-new-deep-architecture-for-reciprocal | 60.1 | 43 |
| retinamask-learning-to-predict-masks-improves | 58.6 | 42.3 |
| deformable-convolutional-networks | 58.0 | - |
| probabilistic-two-stage-detection | 74.0 | 61.6 |
| usb-universal-scale-object-detection | 70.0 | 55.8 |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| nms-strikes-back | 80.4 | 70.2 |
| solq-segmenting-objects-by-learning-queries | - | - |
| 190807919 | 65.9 | 51.2 |
| yolov7-trainable-bag-of-freebies-sets-new | - | - |
| reppoints-v2-verification-meets-regression | 68.9 | 53.4 |
| spinenet-learning-scale-permuted-backbone-for | 62.3 | 46.1 |
| acquisition-of-localization-confidence-for | - | - |
| queryinst-parallelly-supervised-mask-query | 75.9 | 61.9 |
| internimage-exploring-large-scale-vision | - | - |
| pp-yoloe-an-evolved-version-of-yolo | 60.5 | 46.6 |
| foveabox-beyond-anchor-based-object-detector | 63.5 | 47.7 |
| simple-copy-paste-is-a-strong-data | - | - |
| bottom-up-object-detection-by-grouping | 60.5 | 47.0 |
| cornernet-detecting-objects-as-paired | 53.7 | 40.1 |
| pp-yoloe-an-evolved-version-of-yolo | 66.5 | 53.0 |
| 190807919 | - | 48.6 |
| a-ranking-based-balanced-loss-function | 69.3 | 52.5 |
| generalized-focal-loss-v2-learning-reliable | 70.9 | 59.2 |
| cascade-rpn-delving-into-high-quality-region | 59.4 | 43.8 |
| leyolo-new-scalable-and-efficient-cnn | - | - |
| yolox-exceeding-yolo-series-in-2021 | 69.6 | 55.7 |
| florence-a-new-foundation-model-for-computer | - | - |
| cascade-r-cnn-delving-into-high-quality | 62.1 | 46.3 |
| yolov3-an-incremental-improvement | - | - |
| spinenet-learning-scale-permuted-backbone-for | 70.4 | 54.9 |
| softer-nms-rethinking-bounding-box-regression | - | - |
| cascade-r-cnn-high-quality-object-detection | 62.1 | 46.3 |
| ota-optimal-transport-assignment-for-object | 68.6 | 57.1 |
| istr-end-to-end-instance-segmentation-with | - | - |
| rethinking-pre-training-and-self-training | - | - |
| exploring-target-representations-for-masked | - | - |
| nas-fcos-fast-neural-architecture-search-for | - | - |
| an-analysis-of-scale-invariance-in-object-1 | 67.3 | 51.1 |
| grounding-dino-marrying-dino-with-grounded | - | - |
| sniper-efficient-multi-scale-training | 67.0 | 51.6 |
| cbnet-a-novel-composite-backbone-network | 71.9 | 58.5 |
| exploring-target-representations-for-masked | - | - |
| revisiting-the-sibling-head-in-object | 71.9 | 56.0 |
| single-shot-refinement-neural-network-for | 58.7 | 40.8 |
| detectors-detecting-objects-with-recursive-1 | 73.5 | 60.1 |
| pafnet-an-efficient-anchor-free-object | 59.8 | 45.3 |
| ibot-image-bert-pre-training-with-online | - | - |
| centermask-real-time-anchor-free-instance-1 | 63.4 | 48.4 |
| fcos-fully-convolutional-one-stage-object | 60.4 | 45.3 |
| istr-end-to-end-instance-segmentation-with | - | - |
| gradient-harmonized-single-stage-detector | 62.8 | 44.2 |
| multiple-anchor-learning-for-visual-object | - | - |
| hierarchical-shot-detector | 61.2 | 46.9 |
| m2det-a-single-shot-object-detector-based-on | 59.4 | 41.7 |
| spinenet-learning-scale-permuted-backbone-for | 71.8 | 56.5 |
| モデル 259 | 1 | 1 |
| swin-transformer-hierarchical-vision | - | - |
| dynamic-head-unifying-object-detection-heads | 72.1 | 59.3 |
| scale-aware-trident-networks-for-object | 69.7 | 53.5 |