Current browse context:
cs.CV
Change to browse by:
References & Citations
Computer Science > Computer Vision and Pattern Recognition
Title: Modular network for high accuracy object detection
(Submitted on 24 Jan 2020 (v1), last revised 13 Sep 2020 (this version, v3))
Abstract: We present a novel modular object detection convolutional neural network that significantly improves the accuracy of object detection. The network consists of two stages in a hierarchical structure. The first stage is a network that detects general classes. The second stage consists of separate networks to refine the classification and localization of each of the general classes objects. Compared to a state of the art object detection networks the classification error in the modular network is improved by approximately 3-5 times, from 12% to 2.5 %-4.5%. This network is easy to implement and has a 0.94 mAP. The network architecture can be a platform to improve the accuracy of widespread state of the art object detection networks and other kinds of deep learning networks. We show that a deep learning network initialized by transfer learning becomes more accurate as the number of classes it later trained to detect becomes smaller.
Submission history
From: Erez Yahalomi [view email][v1] Fri, 24 Jan 2020 21:38:22 GMT (5139kb,D)
[v2] Sun, 15 Mar 2020 15:50:45 GMT (4952kb)
[v3] Sun, 13 Sep 2020 19:03:31 GMT (1896kb,D)
Link back to: arXiv, form interface, contact.