In the automotive industry, tire texture image classification is crucial for ensuring safety and maintaining quality control. This work presents the surging demand for automated quality control and, more specifically, the critical challenge of automated tire texture image classification. To achieve this, we curated a specialized dataset of 1968 digital tire images, categorized as either defective or in good condition, and augmented it to a final size of 3934 images. We present TireInspectNet, a fully custom-designed Convolutional Neural Network for tire defect detection with utmost precision. It is a very compact-sized model, only 1.32 MB in size, making it 37 times smaller than models available today. Despite its compact size, it achieved an impressive test accuracy of 97.72% on our custom dataset, demonstrating exceptional performance. The model’s small size and high accuracy make it an economically viable solution for deployment in resource-constrained environments, real-time applications on edge devices, and scalability for large-scale applications. TireInspectNet’s performance is strong enough to make it a strong competitor to existing models, demonstrating its great potential for tyre problem identification in the automobile sector.