The chemical treatment of seeds is a fundamental practice that ensures protection against pests and diseases, thus promoting robust plant establishment. However, this process is susceptible to failures, particularly in the form of inadequate coating. As such, the precise assessment of treatment quality emerges as a critical factor in securing high-performance crop yields. In this work, we present an approach based on image processing and convolutional neural networks (CNNs) to segment and predict the quality of chemical coverage on seeds from RGB images. The seeds were arranged on a homogeneous surface and labeled in six categories (C1 to C6), according to the level of chemical coating, with C1 corresponding to no treatment and C6 to adequate treatment, totaling 1,165 seed images, with half of the images captured under natural light and the other half under artificial lighting. For segmentation, granulometric analysis and morphological segmentation techniques were applied, allowing the individual isolation of each seed. For classification, a CNN based on the MobileNetV2 architecture was used, with fine-tuning and data augmentation techniques. The model achieved an average F1 score of 0.96, performing well in all classes. The results demonstrate that the proposed approach is capable of identifying subtle variations in color and uniformity of coverage with excellence, indicating its potential for embedded automated screening applications. The proposal contributes to the standardization and automation of seed evaluation, with direct applicability in the agribusiness sector.