Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 23 min ago
Peer-reviewedCC BY 4.0PLOS One

A spatiotemporal deep learning framework for detecting unusual human activity in surveillance videos

  • Manoj Kumar1
  • Mantosh Biswas2
  • Anoop Kumar Patel3
  • Abhay Kumar4
  • Kumar Abhishek5
  • Ahamed Shafeeq B. M.
5 affiliations
  1. 1School of Computer Science Engineering & Technology, Bennett University, Greater Noida, India
  2. 2Department of Computer Science, North Campus, University of Delhi, Delhi, India
  3. 3National Institute of Technology, Kurukshetra, Haryana, India
  4. 4Department of Computer Science and Engineering, NIT Patna, Patna, Bihar, India
  5. 5Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, India
Open at sourcePDFDOICite

Abstract

With smart gadgets and computers everywhere, real-time video monitoring is now necessary to keep people secure and lower the chance of unexpected human actions. However, current surveillance systems have a lot of trouble effectively spotting unusual actions in real life. Most of the advanced anomaly detection algorithms are trained and tested on synthetic or simulated video datasets. These datasets frequently do not show how complicated and different real-life settings can be. Because of this, their usefulness and capacity to be used in real life are still restricted. We propose a transfer learning-based hybrid deep framework for finding unusual activity in real-world surveillance. We test the model using raw video streams from a subset of the UCF-Crime dataset that shows actual situations of people doing things. First, a pre-trained DenseNet-201, a convolutional neural network, extracts spatial data from video frames that have already been processed. A Bidirectional Long Short-Term Memory (Bi-LSTM) network is used to describe temporal relationships between video sequences in a useful way. The spatial data and the Bi-LSTM’s temporal representations are then integrated to make anomaly identification more precise and dependable. This integrated spatiotemporal architecture identifies abnormal actions of human beings in challenging monitoring situations. The model that was recommended performs better than the state of art methods. It gets 95.04% accuracy on the UCF50 dataset and 62.04% on the untrimmed UCF-Crime dataset. These results suggest that the model performs effective in real-life scenarios.

Related

  1. Localising the Seizure Onset Zone from Routine Interictal and Ictal Scalp EEG Via a Factorised Spatiotemporal Ensemble Deep Learning Framework
    S. Shi
    bioRxivPreprintCC BYFull textNeurosciencebioRxiv
  2. TxCyto: A machine learning framework for estimating cytokine activity from whole transcriptome
    R. Kumar, S. Li, P. Jiang +1
    bioRxivPreprintCC0Full textSystems & synthetic biologybioRxiv
  3. Quantifying Assay Confidence in Barrier Organ-on-Chip Systems: A Monte Carlo Assay-Readiness Framework for Detecting Modest Permeability Shifts
    P. Dama, S. Sahare
    bioRxivPreprintAll rights reservedFull textSystems & synthetic biologybioRxiv
  4. SafeLens: A Statistically Controlled Nonlinear Machine Learning Framework for Characterizing Hidden Dynamics in Event-Locked Neural Activity
    C. B. Minasi, U. Hassan, C. C. Cline +1
    bioRxivPreprintCC BYFull textNeurosciencebioRxiv
  5. Evaluating home-cage-activity metrics for detecting postprocedural pain in a plantar incision mouse model
    M. E. Malmberg, T. Bertelsen, P. Bollen +3
    bioRxivPreprintCC BYFull textZoology & animal behaviourbioRxiv

Cite

BibTeXRIS
BibTeX
@article{kumar2026spatiotemporal,
  title = {{A spatiotemporal deep learning framework for detecting unusual human activity in surveillance videos}},
  author = {Manoj Kumar and Mantosh Biswas and Anoop Kumar Patel and Abhay Kumar and Kumar Abhishek and Ahamed Shafeeq B. M.},
  journal = {PLOS One},
  year = {2026},
  month = sep,
  doi = {10.1371/journal.pone.0355690},
  url = {https://doi.org/10.1371/journal.pone.0355690},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - A spatiotemporal deep learning framework for detecting unusual human activity in surveillance videos
AU  - Manoj Kumar
AU  - Mantosh Biswas
AU  - Anoop Kumar Patel
AU  - Abhay Kumar
AU  - Kumar Abhishek
AU  - Ahamed Shafeeq B. M.
PY  - 2026
DA  - 2026/09/23
JO  - PLOS One
DO  - 10.1371/journal.pone.0355690
UR  - https://doi.org/10.1371/journal.pone.0355690
AB  - With smart gadgets and computers everywhere, real-time video monitoring is now necessary to keep people secure and lower the chance of unexpected human actions. However, current surveillance systems have a lot of trouble effectively spotting unusual actions in real life. Most of the advanced anomaly detection algorithms are trained and tested on synthetic or simulated video datasets. These datasets frequently do not show how complicated and different real-life settings can be. Because of this, their usefulness and capacity to be used in real life are still restricted. We propose a transfer learning-based hybrid deep framework for finding unusual activity in real-world surveillance. We test the model using raw video streams from a subset of the UCF-Crime dataset that shows actual situations of people doing things. First, a pre-trained DenseNet-201, a convolutional neural network, extracts spatial data from video frames that have already been processed. A Bidirectional Long Short-Term Memory (Bi-LSTM) network is used to describe temporal relationships between video sequences in a useful way. The spatial data and the Bi-LSTM’s temporal representations are then integrated to make anomaly identification more precise and dependable. This integrated spatiotemporal architecture identifies abnormal actions of human beings in challenging monitoring situations. The model that was recommended performs better than the state of art methods. It gets 95.04% accuracy on the UCF50 dataset and 62.04% on the untrimmed UCF-Crime dataset. These results suggest that the model performs effective in real-life scenarios.
SN  - 1932-6203
ER  -