文章目录
- Introduction
- 2022
* - SLICING AIDED HYPER INFERENCE AND FINE-TUNING FOR SMALL OBJECT DETECTION
- YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors
Introduction
这里会不定期更新新颖的目标检测方法。
2022
SLICING AIDED HYPER INFERENCE AND FINE-TUNING FOR SMALL OBJECT DETECTION
code: https://paperswithcode.com/paper/slicing-aided-hyper-inference-and-fine-tuning
YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors
code: https:// github.com/WongKinYiu/yolov7
摘要: YOLOv7在5 FPS到160 FPS的速度和精度上都超过了所有已知的目标探测器,在GPU V100的实时目标探测器中具有最高的56.8% AP。YOLOv7-E6目标探测器(56 FPS V100,55.9%美联社)优于变压器探测器SWINL级联面具R-CNN(9.2FPS100,53.9%美联社)速度为509%和2%,和基于卷积探测器ConvNeXt-XL级联面具R-CNN(8.6FPS100,55.2%美联社)速度551%和0.7%的精度,以及YOLOv7超越: YOLOR,YOLOX,规模,YOLOv4,数据,DINO-5比例R50,视频适配器b和许多其他对象探测器的速度和准确性。此外,我们只在MS COCO数据集上开始训练YOLOv7,而没有使用任何其他数据集或预先训练的权重。
Original: https://blog.csdn.net/weixin_42990464/article/details/127358758
Author: 点PY
Title: 目标检测论文、代码、数据集汇总
原创文章受到原创版权保护。转载请注明出处:https://www.johngo689.com/629024/
转载文章受原作者版权保护。转载请注明原作者出处!