A Framework for Anomaly Classification Using Deep Transfer Learning Approach
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Over the last few years, surveillance CCTV cameras have rapidly grown to monitor human activities. Suspicious activities like assault, gun violence, kidnapping need to be observed in public places like malls, public roads, colleges, etc. There is a need for such a surveillance system that automatically recognizes human behavior, such as violent and non-violent actions. Action recognition has become an active research topic for researchers within the computer vision field. However, the human behavior recognition community has mainly focused only on regular actions like walking, running, jogging, etc. Though, detecting behavior in anomaly subjects like assault violence, gun violence, or general aggressive behavior has been comparatively less research in these specific events due to a lack of datasets and algorithms. Thus, there is an increasing demand for datasets to develop abnormal behavior algorithms that can classify anomaly actions. In this paper, the novel dataset is proposed named Human Behavior Dataset 2021 (HBD21). There are four categories of videos available in this dataset: Assault violence, Gun violence, Sabotage violence, and Normal events. This proposed dataset contains a total of 456 videos. Each video has the same length of each category. This paper aims to make a robust surveillance system framework with the help of a deep transfer learning approach and proposed a novel hybrid model. In this view, the current research work is categorized into three phases. Firstly, the preprocessing technique is applied to enhance the brightness of videos, and for resizing then, frames are extracted from each video. Secondly, the transfer learning-based Xception model is used to extract relevant features from frames. The third phase is a classification of behaviors in which a modified LSTM technique is applied. The model is trained using LSTM on the HBD21 dataset. Moreover, using proposed methods on the HBD21 dataset, the accuracy is obtained 97.25% overall.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it