Cover Image

Dynamic Curriculum Learning for Great Ape Detection in the Wild

CC BY

Saved in:
Bibliographic Details
Main Authors: Xinyu, Yang, Tilo, Burghardt, Majid, Mirmehdi
Format: Book
Language:English
Published: Springer 2023
Subjects:
We propose a novel end-to-end curriculum learning approach for sparsely labelled animal datasets leveraging large volumes of unlabelled data to improve supervised species detectors. We exemplify the method in detail on the task of finding great apes in camera trap footage taken in challenging real-world jungle environments. In contrast to previous semi-supervised methods, our approach adjusts learning parameters dynamically over time and gradually improves detection quality by steering training towards virtuous self-reinforcement. To achieve this, we propose integrating pseudo-labelling with curriculum learning policies and show how learning collapse can be avoided. We discuss theoretical arguments, ablations, and significant performance improvements against various state-of-the-art systems when evaluating on the Extended PanAfrican Dataset holding approx. 1.8M frames.
Online Access:https://link.springer.com/article/10.1007/s11263-023-01748-3
https://dlib.phenikaa-uni.edu.vn/handle/PNK/8268
Tags: Add Tag
No Tags, Be the first to tag this record!
id oai:localhost:PNK-8268
record_format dspace
spelling oai:localhost:PNK-82682023-04-25T03:18:01Z Dynamic Curriculum Learning for Great Ape Detection in the Wild Xinyu, Yang Tilo, Burghardt Majid, Mirmehdi We propose a novel end-to-end curriculum learning approach for sparsely labelled animal datasets leveraging large volumes of unlabelled data to improve supervised species detectors. We exemplify the method in detail on the task of finding great apes in camera trap footage taken in challenging real-world jungle environments. In contrast to previous semi-supervised methods, our approach adjusts learning parameters dynamically over time and gradually improves detection quality by steering training towards virtuous self-reinforcement. To achieve this, we propose integrating pseudo-labelling with curriculum learning policies and show how learning collapse can be avoided. We discuss theoretical arguments, ablations, and significant performance improvements against various state-of-the-art systems when evaluating on the Extended PanAfrican Dataset holding approx. 1.8M frames. CC BY We propose a novel end-to-end curriculum learning approach for sparsely labelled animal datasets leveraging large volumes of unlabelled data to improve supervised species detectors. We exemplify the method in detail on the task of finding great apes in camera trap footage taken in challenging real-world jungle environments. In contrast to previous semi-supervised methods, our approach adjusts learning parameters dynamically over time and gradually improves detection quality by steering training towards virtuous self-reinforcement. To achieve this, we propose integrating pseudo-labelling with curriculum learning policies and show how learning collapse can be avoided. We discuss theoretical arguments, ablations, and significant performance improvements against various state-of-the-art systems when evaluating on the Extended PanAfrican Dataset holding approx. 1.8M frames. 2023-04-25T03:18:01Z 2023-04-25T03:18:01Z 2023 Book https://link.springer.com/article/10.1007/s11263-023-01748-3 https://dlib.phenikaa-uni.edu.vn/handle/PNK/8268 en application/pdf Springer
institution Digital Phenikaa
collection Digital Phenikaa
language English
topic We propose a novel end-to-end curriculum learning approach for sparsely labelled animal datasets leveraging large volumes of unlabelled data to improve supervised species detectors. We exemplify the method in detail on the task of finding great apes in camera trap footage taken in challenging real-world jungle environments. In contrast to previous semi-supervised methods, our approach adjusts learning parameters dynamically over time and gradually improves detection quality by steering training towards virtuous self-reinforcement. To achieve this, we propose integrating pseudo-labelling with curriculum learning policies and show how learning collapse can be avoided. We discuss theoretical arguments, ablations, and significant performance improvements against various state-of-the-art systems when evaluating on the Extended PanAfrican Dataset holding approx. 1.8M frames.
spellingShingle We propose a novel end-to-end curriculum learning approach for sparsely labelled animal datasets leveraging large volumes of unlabelled data to improve supervised species detectors. We exemplify the method in detail on the task of finding great apes in camera trap footage taken in challenging real-world jungle environments. In contrast to previous semi-supervised methods, our approach adjusts learning parameters dynamically over time and gradually improves detection quality by steering training towards virtuous self-reinforcement. To achieve this, we propose integrating pseudo-labelling with curriculum learning policies and show how learning collapse can be avoided. We discuss theoretical arguments, ablations, and significant performance improvements against various state-of-the-art systems when evaluating on the Extended PanAfrican Dataset holding approx. 1.8M frames.
Xinyu, Yang
Tilo, Burghardt
Majid, Mirmehdi
Dynamic Curriculum Learning for Great Ape Detection in the Wild
description CC BY
format Book
author Xinyu, Yang
Tilo, Burghardt
Majid, Mirmehdi
author_facet Xinyu, Yang
Tilo, Burghardt
Majid, Mirmehdi
author_sort Xinyu, Yang
title Dynamic Curriculum Learning for Great Ape Detection in the Wild
title_short Dynamic Curriculum Learning for Great Ape Detection in the Wild
title_full Dynamic Curriculum Learning for Great Ape Detection in the Wild
title_fullStr Dynamic Curriculum Learning for Great Ape Detection in the Wild
title_full_unstemmed Dynamic Curriculum Learning for Great Ape Detection in the Wild
title_sort dynamic curriculum learning for great ape detection in the wild
publisher Springer
publishDate 2023
url https://link.springer.com/article/10.1007/s11263-023-01748-3
https://dlib.phenikaa-uni.edu.vn/handle/PNK/8268
_version_ 1764177436458614784
score 8.887836

Similar Items