About this document

OneNet: End-to-End One-Stage Detection by Nhật Hoàng is a document available to read on EtoBox.

This document presents OneNet, a one-stage object detection model that aims to achieve end-to-end detection without post-processing. It discusses how previous one-stage detectors assign labels based only on location cost, which leads to redundant high-confidence boxes and requires non-maximum suppression. The paper proposes Minimum Cost Assignment, which considers both classification cost and location cost to assign a single positive sample per object, with all others labeled negative. When trained with thi

Author

Nhật Hoàng

Language

EN