Object Detection On Coco O

Metrics

Average mAP

Effective Robustness

Results

Performance results of various models on this benchmark

Model Name	Average mAP	Effective Robustness	Paper Title
SSD (VGG-16)	13.6	0.36	SSD: Single Shot MultiBox Detector
GRiT (ViT-H)	42.9	15.72	GRiT: A Generative Region-to-text Transformer for Object Understanding
UniverseNet (R2-101-DCN)	-	1.86	USB: Universal-Scale Object Detection Benchmark
FIBER-B (Swin-B)	33.7	11.43	Coarse-to-Fine Vision-Language Pre-training with Fusion in the Backbone
DyHead (ResNet-50)	19.3	0.16	Dynamic Head: Unifying Object Detection Heads with Attentions
Mask R-CNN (ResNet-50)	17.1	-	Mask R-CNN
VFNet (RX-101-64x4d)	28.0	5.27	VarifocalNet: An IoU-aware Dense Object Detector
DETA (Swin-L)	48.5	20.15	NMS Strikes Back
Mask R-CNN (ResNet-50)	-	-0.11	Mask R-CNN
GCNet (RX-101-32x4d-DCN)	26.0	4.38	GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond
RetinaNet (ResNet-50)	16.6	0.18	Focal Loss for Dense Object Detection
YOLOv7-E6E	32.0	6.42	YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors
YOLOX-X	30.3	7.26	YOLOX: Exceeding YOLO Series in 2021
DINO (Swin-L)	42.1	15.76	DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection
GLIP-L (Swin-L)	48.0	24.89	Grounded Language-Image Pre-training
YOLOX-S	20.6	2.48	YOLOX: Exceeding YOLO Series in 2021
YOLOS-B (ViT-B)	20.0	1.05	You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection
DyHead (Swin-L)	35.3	10.00	Dynamic Head: Unifying Object Detection Heads with Attentions
YOLOv3 (DarkNet-53)	14.8	-0.37	YOLOv3: An Incremental Improvement
RepPointsV2 (RX-101-64x4d-DCN)	24.9	2.7	RepPoints V2: Verification Meets Regression for Object Detection

0 of 45 row(s) selected.