Abstract

Dashcam footage can play a crucial role in the analysis of traffic accidents, offering a detailed record of the events leading up to and during an accident. In this work, we propose a machine learning-based approach for dashcam speed analysis with a focus on traffic accident analysis. Our research project utilizes convolutional neural networks to automatically process dashcam footage and extract speed information and other relevant metrics. This information can then be used to reconstruct the events leading up to an accident and identify potential causes. We evaluate our approach on a dataset of real-world dashcam footage of accidents and regular driving and demonstrate its effectiveness and precision. This approach aims to be a great aid to traffic accident investigators by making dashcam videos even more useful in the analysis process.