Revolutionizing Video Event Detection: An Efficient Approach Utilizing Temporal Convolutional Neural Networks Using Real-time Performance

Automatic detection, localization and interpretation of an unusual event in a sequence of video is a challenging task due to its ambiguous and complex nature. A key challenge with manual detection and sequential analysis lies in the fact that video analysis often demands an immense amount of time and resources. To address this challenge, we have adopted an approach that involved Temporal Convolutional Neural Networks (TCNNs) and Long Short-Term Memory (LSTM) methods. This methodology offers a more efficient way to detect activities in video data. Furthermore, our approach enhances the capture of contextual information within the video, leading to improved results in video content detection. TCNNs, in particular, stand out as they can process input sequences in parallel, resulting in faster training and inference when compared to Recurrent Neural Networks (RNNs) which process data sequentially. To evaluate the effectiveness of these models, we conducted experiments on a manually annotated dataset known as UCF101, which contains 101 different activities. This dataset comprises 1,200 videos representing various actions. Our experimental results demonstrate that TCNNs achieved an accuracy rate of 93.80%, surpassing other existing models are ResNet and Multiclass Support vector machine in overall performance. TCNN, in particular, exhibited exceptional accuracy, indicating its effectiveness in video event detection.