Plug and Play Active Learning for Object Detection

Paper by: Chenhongyi Yang, Lichao Huang, Elliot J. Crowley

This paper discusses a new active learning algorithm, Plug and Play Active Learning (PPAL), which overcomes the difficulties of previous active learning algorithms when it comes to object detection. PPAL consists of

  • Difficulty Calibrated Uncertainty Sampling, in which a category-wise difficulty coefficient that takes both classification and localisation into account to re-weight object uncertainties for uncertainty-based sampling;
  • Category Conditioned Matching Similarity to compute the similarities of multi-instance images as ensembles of their instance similarities.

PPAL is highly generalisable because it makes no change to model architectures or detector training pipelines.

Abstract:

Annotating data for supervised learning is expensive and tedious, and we want to do as little of it as possible. To make the most of a given "annotation budget" we can turn to active learning (AL) which aims to identify the most informative samples in a dataset for annotation. Active learning algorithms are typically uncertainty-based or diversity-based. Both have seen success in image classification, but fall short when it comes to object detection. We hypothesise that this is because: (1) it is difficult to quantify uncertainty for object detection as it consists of both localisation and classification, where some classes are harder to localise, and others are harder to classify; (2) it is difficult to measure similarities for diversity-based AL when images contain different numbers of objects. We propose a two-stage active learning algorithm Plug and Play Active Learning (PPAL) that overcomes these difficulties. It consists of (1) Difficulty Calibrated Uncertainty Sampling, in which we used a category-wise difficulty coefficient that takes both classification and localisation into account to re-weight object uncertainties for uncertainty-based sampling; (2) Category Conditioned Matching Similarity to compute the similarities of multi-instance images as ensembles of their instance similarities. PPAL is highly generalisable because it makes no change to model architectures or detector training pipelines. We benchmark PPAL on the MS-COCO and Pascal VOC datasets using different detector architectures and show that our method outperforms the prior state-of-the-art.

Summary:

Plug and Play Active Learning for Object Detection
Annotating data for supervised learning is expensive and tedious, and we wantto do as little of it as possible. To make the most of a given “annotationbudget” we can turn to active learning (AL) which aims to identify the mostinformative samples in a dataset for annotation. Active learning algori…
arXiv.orgChenhongyi Yang

Code:

GitHub - ChenhongyiYang/PPAL: Plug and Play Active Learning for Object Detection
Plug and Play Active Learning for Object Detection - GitHub - ChenhongyiYang/PPAL: Plug and Play Active Learning for Object Detection
GitHubChenhongyiYang

Do you want to work with us on any of these problems?
We work on similar exciting problems, sometimes for fun and sometimes with commercial aspects. If you have ideas or want to explore these fields with us, please get in touch with us or reply to this email.