Enhancing Anomaly Detection Performance: Deep Learning Models Evaluation
Detection of anomalies within video streams continues to be challenging, mostly due to the complexities involved in distinguishing abnormal activities from normal ones. This study aimed to enhance anomaly detection performance by evaluating different deep learning models and optimizers. Utilizing t...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IIUM Press, International Islamic University Malaysia
2025-05-01
|
| Series: | International Islamic University Malaysia Engineering Journal |
| Subjects: | |
| Online Access: | https://journals.iium.edu.my/ejournal/index.php/iiumej/article/view/3287 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Detection of anomalies within video streams continues to be challenging, mostly due to the complexities involved in distinguishing abnormal activities from normal ones. This study aimed to enhance anomaly detection performance by evaluating different deep learning models and optimizers. Utilizing the Keras framework and Python on a Kaggle notebook, the experiment explored the effectiveness of DenseNet121, VGG19, ResNet50, and InceptionV3 models in conjunction with Adam, SGD, RMSprop, and Adagrad optimizers. A UCF Crimes dataset subset focused on Accuracy, F1 Score, and AUC evaluation metrics. The results establish that the InceptionV3 model paired with the Adam optimizer outperforms the other combinations, attaining AUC scores of 0.9918. In contrast to other state-of-the-art models such as DenseNet121 and ResNet50, InceptionV3 presents enhanced precision and adaptability in handling the variability found in video anomaly datasets. This study enhances security by providing insights into enhanced model-optimizer combinations, advancing video surveillance approaches, and providing support for developing robust anomaly detection systems.
ABSTRAK: Pengesanan anomali dalam strim video terus mencabar, kebanyakan disebabkan oleh kerumitan yang terlibat dalam membezakan aktiviti tidak normal dari biasa. Kajian ini cuba meningkatkan prestasi pengesanan anomali dengan menilai model dan pengoptimum pembelajaran mendalam yang berbeza. Menggunakan rangka kerja Keras dan Python pada komputer riba Kaggle, eksperimen ini meneroka keberkesanan model DenseNet121, VGG19, ResNet50 dan InceptionV3 bersama pengoptimum Adam, SGD, RMSprop dan Adagrad. Subset data Jenayah UCF digunakan, memfokuskan pada ketepatan, Skor F1 dan metrik penilaian AUC. Dapatan kajian menunjukkan bahawa model InceptionV3 bersama pengoptimum Adam, mengatasi kombinasi lain, mencapai skor AUC 0.9918. Berbeza dengan model canggih lain seperti DenseNet121 dan ResNet50, InceptionV3 mempunyai ketepatan dan kebolehsuaian yang tinggi dalam mengendalikan kebolehubahan yang terdapat dalam set data anomali video. Kajian ini menyumbang kepada peningkatan keselamatan dengan memberi gabungan pengoptimum bersama model yang dipertingkatkan, memajukan pendekatan pengawasan video dan menyediakan sokongan bagi pembangunan sistem pengesanan anomali yang teguh.
|
|---|---|
| ISSN: | 1511-788X 2289-7860 |