Redirigiendo al acceso original de articulo en 24 segundos...
Inicio  /  Applied Sciences  /  Vol: 12 Par: 20 (2022)  /  Artículo
ARTÍCULO
TITULO

FastDARTSDet: Fast Differentiable Architecture Joint Search on Backbone and FPN for Object Detection

Chunxian Wang    
Xiaoxing Wang    
Yiwen Wang    
Shengchao Hu    
Hongyang Chen    
Xuehai Gu    
Junchi Yan and Tao He    

Resumen

Neural architecture search (NAS) is a popular branch of automatic machine learning (AutoML), which aims to search for efficient network structures. Many prior works have explored a wide range of search algorithms for classification tasks, and have achieved better performance than manually designed network architectures. However, few works have explored NAS for object detection tasks due to the difficulty to train convolution neural networks from scratch. In this paper, we propose a framework, named as FastDARTSDet, to directly search on a larger-scale object detection dataset (MS-COCO). Specifically, we propose to apply differentiable architecture search method (DARTS) to jointly search backbone and feature pyramid network (FPN) architectures for object detection task. Extensive experimental results on MS-COCO show the efficient and efficacy of our method. Specifically, our method achieves 40.0% mean average precision (mAP) on the test set, outperforming many recent NAS methods.

 Artículos similares

       
 
Zheng Zhao, Jialing Yuan and Luhao Chen    
Air Traffic Flow Management (ATFM) delay can quantitatively reflect the congestion caused by the imbalance between capacity and demand in an airspace network. Furthermore, it is an important parameter for the ex-post analysis of airspace congestion and t... ver más
Revista: Aerospace

 
Nikola Andelic and Sandi Baressi ?egota    
This investigation underscores the paramount imperative of discerning network intrusions as a pivotal measure to fortify digital systems and shield sensitive data from unauthorized access, manipulation, and potential compromise. The principal aim of this... ver más
Revista: Information

 
Danilo Pau, Andrea Pisani and Antonio Candelieri    
In the context of TinyML, many research efforts have been devoted to designing forward topologies to support On-Device Learning. Reaching this target would bring numerous advantages, including reductions in latency and computational complexity, stronger ... ver más
Revista: Algorithms

 
Ulrich A. Ngamalieu-Nengoue, Pedro L. Iglesias-Rey, F. Javier Martínez-Solano and Daniel Mora-Meliá    
Extreme rainfall events cause immense damage in cities where drainage networks are nonexistent or deficient and thus unable to transport rainwater. Infrastructure adaptations can reduce flooding and help the population avoid the associated negative conse... ver más
Revista: Water

 
Konstantin Gaipov, Daniil Tausnev, Sergey Khodenkov, Natalya Shepeta, Dmitry Malyshev, Aleksey Popov and Lev Kazakovtsev    
Rapid growth in the volume of transmitted information has lead to the emergence of new wireless networking technologies with variable heterogeneous topologies. With limited radio frequency resources, optimal routing problems arise, both at the network de... ver más
Revista: Algorithms