Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Jun 13, 2026

Decoding Natural Behavior from Neuroethological Embedding
08:00

Decoding Natural Behavior from Neuroethological Embedding

Published on: October 3, 2025

PMTNet: A Part-Centric Missing-Aware Temporal Network for Cat Behavior Recognition in Unconstrained Videos.

Chunxi Tu1, Jiatao Wu1, Zeguang Huang2

  • 1College of Artificial Intelligence and Low-Altitude Technology, South China Agricultural University, Guangzhou 510642, China.

Animals : an Open Access Journal From MDPI
|June 12, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Multi-omics machine learning-driven investigation of super-enhancers signatures and prognostic biomarkers in lung adenocarcinoma.

Functional & integrative genomics·2026
Same author

Design and implementation of a thermoelectric cooler-based temperature control system for the resonator tube wall of an optically pumped terahertz gas laser.

The Review of scientific instruments·2026
Same author

Alkaline selective separation of silicon/vanadium from vanadium-bearing shale and recovery of silica.

Scientific reports·2026
Same author

Mechanically Adaptive Polyimide Interfaces for Stable High-Voltage NCM-Sulfide All-Solid-State Batteries.

Advanced science (Weinheim, Baden-Wurttemberg, Germany)·2026
Same author

On the interpretation of negative efficacy in the MUCOSA trial: is the role of eosinophils in refractory chronic cough truly dismissed?

The European respiratory journal·2026
Same author

Genetic Diversity and Core Germplasm Development in Yunnan Tartary Buckwheat Based on Phenotypic and SNP Data.

Plants (Basel, Switzerland)·2026

This study introduces PMTNet, a novel part-centric temporal network that enhances cat behavior recognition in videos, even with challenging, intermittent visibility of body parts like the head and tail.

Area of Science:

  • Computer Vision
  • Animal Behavior Analysis
  • Machine Learning

Background:

  • Recognizing cat behavior in real-world videos is crucial for animal welfare and veterinary care.
  • Challenges arise from deformable and intermittently visible body parts (head, tail), hindering accurate analysis.
  • Existing methods struggle with the dynamic and unconstrained nature of video data.

Purpose of the Study:

  • To improve clip-level cat behavior recognition in unconstrained videos, specifically addressing unstable part visibility.
  • To develop and evaluate a part-centric temporal network (PMTNet) for robust cat behavior analysis.
  • To enhance the accuracy of automated systems for monitoring feline welfare and health.

Main Methods:

  • Proposed PMTNet, a part-centric temporal network utilizing DEIM-based detection for cat body, head, and tail.
Keywords:
DEIM-based detectionanimal behavior analysiscat behavior recognitionmissing-aware fusionpart-centric temporal modeling

More Related Videos

Tracking Rats in Operant Conditioning Chambers Using a Versatile Homemade Video Camera and DeepLabCut
08:32

Tracking Rats in Operant Conditioning Chambers Using a Versatile Homemade Video Camera and DeepLabCut

Published on: June 15, 2020

Related Experiment Videos

Last Updated: Jun 13, 2026

Decoding Natural Behavior from Neuroethological Embedding
08:00

Decoding Natural Behavior from Neuroethological Embedding

Published on: October 3, 2025

Tracking Rats in Operant Conditioning Chambers Using a Versatile Homemade Video Camera and DeepLabCut
08:32

Tracking Rats in Operant Conditioning Chambers Using a Versatile Homemade Video Camera and DeepLabCut

Published on: June 15, 2020

  • Implemented a dynamic detector selection strategy based on video-domain continuity and stability.
  • Modeled behavior using Region of Interest (ROI) appearance features and explicit geometric motion cues.

    • Main Results:

    • PMTNet achieved 93.1% Top-1 Accuracy and 90.9% Macro-F1 on a cat behavior dataset.
    • Ablation studies confirmed the contributions of detector choice, complementary part cues, and missing-aware fusion.
    • PMTNet outperformed representative end-to-end video recognition baselines on the evaluated dataset.

    Conclusions:

    • Part-centric temporal modeling effectively addresses challenges in cat behavior recognition under unstable part visibility.
    • PMTNet demonstrates significant potential for real-world applications in animal welfare monitoring and veterinary assessment.
    • The proposed framework offers a robust solution for analyzing feline behavior in unconstrained video environments.