目前常见的目标检测方法及技巧收藏整理

一、目标检测方法

Anchor

1. One Stage

Yolo Family：yolov1...yolov3...yolov5
SSD Family
RetinaNet
M2Det

2. Two Stage

RCNN
Fast RCNN
Faster RCNN
Mask RCNN
RFCN
Light-Head-RCNN
Cascade RCNN
Libra RCNN
DetectoRS

Anchor Free

ConnerNet
Foveabox
CenterNet
FCOS
RepPoint
ATSS
FreeAnchor
FSAF

Transformer

DETR

NAS

DetNAS
NAS-FPN
Auto-FPN
NAS-FCOS
SpineNet
Hit-Detector
EfficientDet

Semi-supervised learning

STAC

二、Tricks

Data Augmentation

SSD augmentation
Random erase
Mixup
Mosaic

Anchor

Guided Anchor

NMS

Soft-NMS

Loss

1. Classification 分类

Focal loss 焦点损失
Generalized Focal loss 广义焦点损失

2. Regression 回归

IoU
GIoU
DIoU
CIoU

Imbalance

Class
Scale
Spatial
Objective

参考文章

本文来自互联网用户投稿，文章观点仅代表作者本人，不代表本站立场，不承担相关法律责任。如若转载，请注明出处。 如若内容造成侵权/违法违规/事实不符，请点击【内容举报】进行投诉反馈！