A Novel Survey on Image Classification Models for Explainable Predictions using Computer Vision

In the field of image classification, while algorithms perform well at recognizing objects, understanding their decision-making processes remains challenging and limiting the applicability in critical domains. This research study analyzes the evolution of image classification models aimed at bridging the gap between prediction and explanation. This research study evaluates various techniques, encompassing rule-based models, decision trees, and local approximation methods like LIME, which offer insights into the factors influencing the proposed model’s decision-making. Additionally, the saliency maps visually interpret image regions crucial for prediction comprehension. Nonetheless, the model development encounters challenges, requiring a balance between accuracy and interpretability, often necessitating trade-offs amidst complexity considerations.