Towards Real-Time Explainable AI: Using Class Activation Mapping for Brake Prediction in YOLO-Based Systems - Figure 5

Figure 5 displays the training and validation loss curves for the trained black-box unified model. The plot demonstrates how the model's performance evolves over the course of training, highlighting the reduction in loss values across epochs. This reduction suggests that the model is increasingly adapting to the data, with both training and validation losses exhibiting a steady convergence trend.