[2005.09973] Dynamic Refinement Network for Oriented and Densely Packed Object Detectionopen searchopen navigation menucontact arXivsubscribe to arXiv mailings

Object detection has achieved remarkable progress in the past decade. However, the detection of oriented and densely packed objects remains challenging because of following inherent reasons: (1) receptive fields of neurons are all axis-aligned and of the same shape, whereas objects are usually of diverse shapes and align along various directions; (2) detection models are typically trained with generic knowledge and may not generalize well to handle specific objects at test time; (3) the limited dataset hinders the development on this task. To resolve the first two issues, we present a dynamic refinement network that consists of two novel components, i.e., a feature selection module (FSM) and a dynamic refinement head (DRH). Our FSM enables neurons to adjust receptive fields in accordance with the shapes and orientations of target objects, whereas the DRH empowers our model to refine the prediction dynamically in an object-aware manner. To address the limited availability of related ben

2 mentions: @AkiraTOSEI@AkiraTOSEI
Keywords: 物体検出
Date:

Referring Tweets

@AkiraTOSEI
@AkiraTOSEI t.co/yUJQwruf1u 密で軸線が画像に対して回転している物体検出は難しいが、BoundingBoxを使わず物体中心を回帰するCenterNetをベースに、検出対象の線にそって検出ができるようにしたモデルを提案。Deformable Convを使って回転に対応し、フィルターを生成することでサンプル毎に動的に対応 t.co/gFFzCYLWG2
@AkiraTOSEI
@AkiraTOSEI t.co/yUJQwruf1u It is difficult to detect objects that are dense and rotated with respect to the image. To overcome this difficulty, they proposed a model to detect along the line of the object using Deformable Conv, and DRH, which dynamically responds to each sample. t.co/q15xe2hBe2

Related Entries

[2011.00147] Pixel-Level Cycle Association: A New Perspective for Domain Adaptive Semantic Segmentat...
Read more [2011.00147] Pixel-Level Cycle Association: A New Perspective for Domain Adaptive Semantic Segmentat...
0 users, 2 mentions 2020/11/11 12:52
[2010.04456] Augmenting Physical Models with Deep Networks for Complex Dynamics Forecasting
Read more [2010.04456] Augmenting Physical Models with Deep Networks for Complex Dynamics Forecasting
0 users, 2 mentions 2021/02/12 11:21
[2104.12807] Multimodal Self-Supervised Learning of General Audio Representations
Read more [2104.12807] Multimodal Self-Supervised Learning of General Audio Representations
0 users, 2 mentions 2021/05/28 12:18
[2111.09886v1] SimMIM: A Simple Framework for Masked Image Modeling
Read more [2111.09886v1] SimMIM: A Simple Framework for Masked Image Modeling
0 users, 3 mentions 2021/11/25 01:37

ML-Newsについて

機械学習の技術に関する情報は流速も早いし、分野も多様でキャッチアップが大変です。Twitterで機械学習用のリストを作っても、普段は機械学習以外の話題が多く流れており、効率的に情報収集するのは困難です。

ML-NewsはSNSを情報源とした機械学習に特化したニュースサイトです。機械学習に関する論文ブログライブラリコンペティション発表資料勉強会などの最新の情報を効率的に収集できます。

機械学習を応用した自然言語処理、画像認識、情報検索などの分野の情報や機械学習で必要になるデータ基盤やMLOpsの話題もカバーしています。
安定したサイト運営のためにGitHub sponsorを募集しています。

お知らせ

  • 2021/12/31: デザインを刷新しました
  • 2021/04/08: 日本語Kaggleのカテゴリを新設しました