Like all types of pollution, illegal garbage dumping can have a negative impact on both the environment and human health. Garbage classification is the focus of most recent research. However, there aren’t many works that acknowledge littering behavior. In this paper, we introduce a unique method for detecting unlawful garbage disposal in on-site camera surveillance data. The integrated model detects the illegal littering. The neural networks’ convolutional processes are good at extracting local information, but they have trouble capturing global representations. Vision Transformer’s multi-head self-attention can capture feature dependencies over long distances, but it can also obliterate local feature details. As a result, we propose a new approach that combines the advantages of CNNs and Vision Transformer, based on MobileNet-v2 and Vision Transformer. In addition, our approach leverages temporal channel attention to enhance the model’s ability to interact with temporal input. To confirm the efficacy of the proposed method, we ran tests on our own dataset. With a validation accuracy of 93.64%, the experimental findings show that our model performs well, performing on par with or even better than the existing techniques.