This paper presents an efficient method for estimating gravel particle size distribution using image processing and artificial neural networks (IPNN). Traditional sieve analysis, commonly used for coarse-grained soils, is often time-consuming and labor-intensive, particularly in large infrastructure projects. The proposed approach applies particle boundary detection and shape feature extraction to train a neural network model capable of predicting the grain size distribution curve. Results show strong agreement between the IPNN method and conventional sieve analysis, with a maximum difference of 3.70% in passing percentages for gravel samples. For crushed stone samples, the maximum difference reached 10.90%, mainly for larger particles. Overall, the technique offers a promising and practical alternative for material quality control in large-scale projects.