Proceedings of the 14th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Application 2019
DOI: 10.5220/0007248601810190
|View full text |Cite
|
Sign up to set email alerts
|

Active Learning for Deep Object Detection

Abstract: The great success that deep models have achieved in the past is mainly owed to large amounts of labeled training data. However, the acquisition of labeled data for new tasks aside from existing benchmarks is both challenging and costly. Active learning can make the process of labeling new data more efficient by selecting unlabeled samples which, when labeled, are expected to improve the model the most. In this paper, we combine a novel method of active learning for object detection with an incremental learning… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
40
0
1

Year Published

2019
2019
2022
2022

Publication Types

Select...
4
4
1

Relationship

1
8

Authors

Journals

citations
Cited by 52 publications
(41 citation statements)
references
References 31 publications
(51 reference statements)
0
40
0
1
Order By: Relevance
“…The first works that handled active learning for large scale object detection [26] used as active learning criterion the simple margin selection method for SVMs [27], which seeks points that most reduce the version space. More recently, methods rely on modern object detectors [28] [29], but still are based on uncertainty indicators like least confidence or 1-vs-2 [30] [31]. Notice that object detection is very close to the instance segmentation task addressed in this work.…”
Section: Related Workmentioning
confidence: 98%
“…The first works that handled active learning for large scale object detection [26] used as active learning criterion the simple margin selection method for SVMs [27], which seeks points that most reduce the version space. More recently, methods rely on modern object detectors [28] [29], but still are based on uncertainty indicators like least confidence or 1-vs-2 [30] [31]. Notice that object detection is very close to the instance segmentation task addressed in this work.…”
Section: Related Workmentioning
confidence: 98%
“…For each image, margin is chosen to be the summation of margins across all the predicted bounding boxes in the image. This is taken from Brust et al [18]. • Entropy (ent): Samples with high entropy in the probability distribution of the predictions are selected.…”
Section: Comparison With Baselinesmentioning
confidence: 99%
“…Many computer vision studies are also underway to minimize the amount of training data and human effort while maximizing the model performance. For example, Roy et al (2018) and Brust et al (2019) employed active learning algorithms to train an object detector. As active learning algorithms select the most informative-to-learn instances from abundant unlabeled training data and train a deep learning model with the selected data first, it is possible to significantly reduce the amount of training data and the human effort required for DB development (Kim et al, 2020b).…”
Section: Database-free Vision-based Monitoringmentioning
confidence: 99%