Intelligent Tyre Condition Assessment Through DL Models for Enhanced Predictive Accuracy

Accurate classification of the condition of tyres is important for the safety and proper maintenance of a vehicle; however, it is difficult since tyres often arrive with intricate features and additional detailed reviews are required. In many cases, traditional techniques fail to reach high accuracy levels, particularly when applying them to large complex data sets. Based on this, deep learning models such as ResNet50, DenseNet121, and MobileNet were implemented and used in testing on tyre classification tasks. The results indicate that DenseNet121 had the highest accuracy at 96.56%, while MobileNet had 95.68%, and ResNet50 had 76.06%. Consequently, the reason for the better performance of DenseNet121 is its dense connectivity architecture, which helps to efficiently reuse features and enables an improved gradient flow, which leads to fast convergence and better generalization. This work demonstrates the potential efficiency of the DenseNet121 model in terms of tyre classification, especially given the intricacy of the datasets considered. Future work will involve experimenting with the model over new and diversified sets to check its generality and to develop further optimizations for practical use.