1、基础
自己对于YOLOV1,2,3都比较熟悉。 RCNN也比较熟悉。这个是自己目前掌握的基础
2、第一步
看一下2019年的井喷的anchor free的网络
3、第二步
看一下以往,引用多的网路
4、第三步
看一下,2020最新的,但是在pwcode上面排名靠前的网络
2020优秀论文:
EfficientDet: Scalable and Efficient Object Detection,57
DetectoRS: Detecting Objects with Recursive Feature Pyramid and Switchable Atrous Convolution
Rethinking Pre-training and Self-training
Revisiting the Sibling Head in Object Detector
Deep High-Resolution Representation Learning for Visual Recognition*52
HoughNet: Integrating near and long-range evidence for bottom-up object detection ??
M2Det: A Single-Shot Object Detector based on Multi-Level Feature Pyramid Network,101
优秀论文:
Acquisition of Localization Confidence for Accurate Object Detection,182
Scale-Aware Trident Networks for Object Detection,141
GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond,107
Path Aggregation Network for Instance Segmentation,400
Hybrid Task Cascade for Instance Segmentation ,139
SNIPER: Efficient Multi-Scale Training,141
Deformable ConvNets v2: More Deformable, Better Results,188
An Analysis of Scale Invariance in Object Detection - SNIP,218
Single-Shot Refinement Neural Network for Object Detection,414
Attention Augmented Convolutional Networks,58
SaccadeNet: A Fast and Accurate Object Detector
two stage:
Grid R-CNN, plus,46
Cascade R-CNN: Delving into High Quality Object Detection,571
Libra R-CNN: Towards Balanced Learning for Object Detection,115
Dynamic R-CNN: Towards High Quality Object Detection via Dynamic Training
one stage:
ExtremNet :Bottom-up Object Detection by Grouping Extreme and Center Points 2019,129
YoLOV5
YOLOv4: Optimal Speed and Accuracy of Object Detection*,17
anchor free的网络:
Scale-Equalizing Pyramid Convolution for Object Detection,2020
FoveaBox: Beyond Anchor-based Object Detector 2019,52
CornerNet: Detecting Objects as Paired Keypoints 2018 ,445
CornerNet-Lite (2019) (https://arxiv.org/abs/1904.08900)
CenterNet: Keypoint Triplets for Object Detection 2019,129
FCOS: Fully Convolutional One-Stage Object Detection 2019,206
Feature Selective Anchor-Free Module for Single-Shot Object Detection,144