A Vision-Based Approach to Enhance Fall Detection with Fine-Tuned Faster R-CNN

Falls are a significant health concern that must be reported immediately. The elderly population and mentally ill individuals require constant care and home assistance. Vision-based fall detection systems (Basavaraj & Kusagur, 2017) provide a dynamic solution to the problem of detecting falls and mitigating the consequences of falls. This paper presents a vision-based fall detection approach that utilizes the CAUCA Fall Dataset (Guerrero et al., 2022) and fine-tunes the Faster R-CNN model with COCO weights to enhance the performance of the fall detection model. We describe the methodology involved in adapting the model while leveraging the rich feature representations learned from the COCO dataset. Extensive experiments are conducted to evaluate the model's performance, including accuracy, precision, recall, and F1-score, on various fall and non-fall scenarios. The results demonstrate the effectiveness of our vision-based fall detection system and its applicability to real-world scenarios.