Object Detection On Coco

评估指标

AP50

AP75

评测结果

各个模型在此基准测试上的表现结果

模型名称	AP50	AP75	Paper Title	Repository
DyHead (ResNet-101)	64.5	50.7	Dynamic Head: Unifying Object Detection Heads with Attentions	-
YOLOv7-D6 (44 fps)	-	-	YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors	-
EfficientDet-D7 (1536)	71.6	56.9	EfficientDet: Scalable and Efficient Object Detection	-
GFLV2 (ResNeXt-101, 32x4d, DCN)	67.6	53.5	Generalized Focal Loss V2: Learning Reliable Localization Quality Estimation for Dense Object Detection	-
Faster R-CNN (ImageNet+300M)	58	40.1	Revisiting Unreasonable Effectiveness of Data in Deep Learning Era	-
GLIP (Swin-L, multi-scale)	79.5	67.7	Grounded Language-Image Pre-training	-
Group DETR v2	81.8	71.1	Group DETR v2: Strong Object Detector with Encoder-Decoder Pretraining	-
Mask R-CNN (ResNeXt-101-FPN)	62.3	43.4	Mask R-CNN	-
ISTR (ResNet50-FPN-3x, single-scale)	-	-	ISTR: End-to-End Instance Segmentation with Transformers	-
CPNDet (Hourglass-104, multi-scale)	67.3	53.7	Corner Proposal Network for Anchor-free, Two-stage Object Detection	-
M2Det (VGG-16, multi-scale)	64.6	49.3	M2Det: A Single-Shot Object Detector based on Multi-Level Feature Pyramid Network	-
YOLOv7-E6 (56 fps)	-	-	YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors	-
A2MIM (ViT-B)	-	-	Architecture-Agnostic Masked Image Modeling -- From ViT back to CNN	-
D-RFCN + SNIP (ResNet-101, multi-scale)	65.5	48.4	An Analysis of Scale Invariance in Object Detection - SNIP	-
LeYOLO-nano@480	-	-	LeYOLO, New Embedded Architecture for Object Detection	-
Centermask + ResNet101	61.6	46.9	CenterMask : Real-Time Anchor-Free Instance Segmentation	-
AC-FPN Cascade R-CNN (X-152-32x8d-FPN-IN5k, multi scale, only CEM)	70.4	57	Attention-guided Context Feature Pyramid Network for Object Detection	-
MnasFPN (MNASNet-B1)	-	-	MnasFPN: Learning Latency-aware Pyramid Architecture for Object Detection on Mobile Devices	-
Gaussian-FCOS	-	-	Localization Uncertainty Estimation for Anchor-Free Object Detection	-
Faster R-CNN	-	-	Speed/accuracy trade-offs for modern convolutional object detectors	-

0 of 262 row(s) selected.