摘要:
paper: "Learning Lightweight Lane Detection CNNs by Self Attention Distillation" code: "https://github.com/cardwing/Codes for Lane Detection" Abstract 阅读全文
摘要:
paper: Objects as Points code: https://github.com/xingyizhou/CenterNet Abstract CenterNet是anchor-free的目标检测方法,用预测出的点表示一个object,同时回归出目标的size和offset。其性能表 阅读全文
摘要:
paper: "Focal Loss for Dense Object Detection" Abstract 1. 目前主流的目标检测方法,one stage相较于two stage的一个痛点就是样本分布的不平衡(目标和背景),导致one stage虽然速度上有优势,但是精度却有所降低。因此,本文 阅读全文
摘要:
Yolo V1 paper: "You Only Look Once: Unified, Real Time Object Detection" slide: "官方slide介绍,很详细" Abstract 1. 如文章名字所述,其完成一次目标检测,只需要look once,也就是不像两阶段的网络 阅读全文
摘要:
paper: "EfficientDet: Scalable and Efficient Object Detection" code "官方tf" "民间PyTroch" Abstract 1. 谷歌大脑团队提出的新的目标检测算法,性能极佳,EfficientDet D0的mAP就超过了yolov 阅读全文