(CVPR2023-IOCFormer)Indiscernible Object Counting in Underwater Scenes 论文阅读

date
Jun 3, 2023
Last edited time
Jun 3, 2023 01:40 PM
status
Published
slug
IOCFormer论文阅读
tags
DL
CV
summary
type
Post
origin
https://www.notion.so/lazurite/CVPR2023-Indiscernible-Object-Counting-in-Underwater-Scenes-1c19a2e28c53486ea4dcd296ea658cd1?pvs=4
Field
Plat
Indiscernible Object Counting in Underwater Scenes
Recently, indiscernible scene understanding has attracted a lot of attention in the vision community. We further advance the frontier of this field by systematically studying a new challenge named...
Indiscernible Object Counting in Underwater Scenes
http://arxiv.org/abs/2304.11677
Indiscernible Object Counting in Underwater Scenes
Sun et al. - 2023 - Indiscernible Object Counting in Underwater Scenes.pdf
8446.4KB
GitHub - GuoleiSun/Indiscernible-Object-Counting: CVPR 2023
CVPR 2023. Contribute to GuoleiSun/Indiscernible-Object-Counting development by creating an account on GitHub.
GitHub - GuoleiSun/Indiscernible-Object-Counting: CVPR 2023
https://github.com/GuoleiSun/Indiscernible-Object-Counting
GitHub - GuoleiSun/Indiscernible-Object-Counting: CVPR 2023
notion image
这篇文章提出了一个名为IOCfish5K的大规模数据集，其中包含了5637张高分辨率的水下场景图像和659,024个注释的中心点。该数据集主要用于解决在难以区分的场景中计数物体的问题。
notion image
本文提出了一种名为IOCFormer的方法，用于在难以区分的场景中计数物体。相比之前的方法，IOCFormer具有以下两个创新点：
  1. 将密度和回归分支结合到统一框架中，从而改善底层特征。
  1. 利用基于密度的方法估计对象密度水平的能力来使（难以辨认的）前景对象脱颖而出并提高基于回归的方法的性能。
notion image
notion image
notion image
notion image
notion image

© Lazurite 2021 - 2025