Related Experiment Video
Updated: Jul 16, 2026
06:57
Pavlovian Conditioned Approach Training in Rats
Published on: February 4, 2016
TP-CanineNet: Temporal Context Contrastive Learning with Pseudo-Label Supervision for Abnormal Behavior Detection of
Xiangyun Guo1, Xiaoya Kong1, Chuiyu Kong2
1College of Management Science and Engineering, Beijing Information Science & Technology University, Beijing 102206, China.
Animals : an Open Access Journal From MDPI
|July 15, 2026
Summary
This study introduces TP-CanineNet, a novel AI model for detecting abnormal dog behaviors when left alone. It improves recognition accuracy for canine distress signals, enhancing pet welfare.
Area of Science:
- Animal Behavior and Welfare
- Computer Vision
- Artificial Intelligence
Background:
- Canine behavioral abnormalities like excessive barking and destructive actions are common when dogs are left alone.
- Current behavior recognition methods lack accuracy due to the complex temporal dynamics of these abnormal behaviors.
Purpose of the Study:
- To develop an advanced model for accurate recognition of canine behavioral abnormalities using video data.
- To improve the detection of temporal features in dog behaviors for better welfare interventions.
Main Methods:
- Developed TP-CanineNet, a model using Weakly Supervised Video Anomaly Detection (WS-VAD).
- Integrated Temporal Context Aggregation (TCA) for temporal dependency and noise suppression.
- Employed Pseudo-Instance Discriminative Enhancement (PIDE) to differentiate abnormal from normal behaviors.
- Utilized a custom Alone-Dog dataset (430 videos) for validation.
Main Results:
- TP-CanineNet achieved a frame-level AUC of 85.19% and an AP of 72.55%.
- Demonstrated significant improvements of 2.20% (AUC) and 8.33% (AP) over baseline models.
- Validated the model's effectiveness on the Alone-Dog dataset.
Conclusions:
- The TP-CanineNet model offers a robust solution for intelligent detection of canine behaviors when unsupervised.
- This technology can aid in early identification of distress, leading to timely interventions and improved canine welfare.
