Indoor Scene Multi-Object Tracking based on Region Search and Memory Buffer Pool

Pattern Recognition, Journal Year: 2025, Volume and Issue: unknown, P. 111623 - 111623

Published: March 1, 2025

Language: Английский

---

The MacqD deep-learning-based model for automatic detection of socially housed laboratory macaques

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 7, 2025

Abstract Despite advancements in video-based behaviour analysis and detection models for various species, existing methods are suboptimal to detect macaques complex laboratory environments. To address this gap, we present MacqD, a modified Mask R-CNN model incorporating SWIN transformer backbone enhanced attention-based feature extraction. MacqD robustly detects their home-cage under challenging scenarios, including occlusions, glass reflections, overexposure light. evaluate compare its performance against pre-existing macaque models, collected analysed video frames from 20 caged rhesus at Newcastle University, UK. Our results demonstrate MacqD’s superiority, achieving median F1-score of 99% with single the focal cage (surpassing next-best by 21%) 90% two macaques. Generalisation tests on different set same animal facility yielded F1-scores 95% 15%) 81% alternative approach 39% ). Finally, was applied videos paired another resulted 90%, reflecting strong generalisation capacity. This study highlights effectiveness accurately detecting across diverse settings.

Language: Английский

---

Computer vision for primate behavior analysis in the wild

Nature Methods, Journal Year: 2025, Volume and Issue: unknown

Published: April 10, 2025

Language: Английский

---

PolarBearVidID: A Video-Based Re-Identification Benchmark Dataset for Polar Bears

Animals, Journal Year: 2023, Volume and Issue: 13(5), P. 801 - 801

Published: Feb. 23, 2023

Automated monitoring systems have become increasingly important for zoological institutions in the study of their animals' behavior. One crucial processing step such a system is re-identification individuals when using multiple cameras. Deep learning approaches standard methodology this task. Especially video-based methods promise to achieve good performance re-identification, as they can leverage movement an animal additional feature. This especially applications zoos, where one has overcome specific challenges changing lighting conditions, occlusions or low image resolutions. However, large amounts labeled data are needed train deep model. We provide extensively annotated dataset including 13 individual polar bears shown 1431 sequences, which equivalent 138,363 images. PolarBearVidID first non-human species date. Unlike typical human benchmark datasets, were filmed range unconstrained poses and conditions. Additionally, approach trained tested on dataset. The results show that animals be identified with rank-1 accuracy 96.6%. thereby characteristic feature it utilized re-identification.

Language: Английский

---

A systematic literature review of visual feature learning: deep learning techniques, applications, challenges and future directions

Multimedia Tools and Applications, Journal Year: 2024, Volume and Issue: unknown

Published: July 20, 2024

Language: Английский

---