The agricultural sector plays a pivotal role in sustaining economies and populations by ensuring food security. Timely and accurate detection of plant diseases is critical to minimizing economic losses and preserving food quality. This study investigates the relationship between the depth of Residual Networks and their classification accuracy in the context of crop disease detection. Four variants of the ResNet family, ranging from ResNet-29 to ResNet-152 are evaluated in order to determine how the number of layers impacts performance. Using a dataset of 54,303 labeled images of crop diseases, the models were trained and assessed for their classification accuracy. Experimental results indicate a positive correlation between model depth and performance, with F1 scores of 0.95 for ResNet-29 that reaches a value of 0.97 for ResNet-152. Based on these findings, the ResNet-152 architecture is recommended for crop disease classification due to its superior accuracy and robustness. This study highlights the importance of model depth in enhancing classification accuracy for agricultural applications.