It can be observed that dynamic developments in artificial intelligence contributing to the evolution of existing techniques used in agricultural research and crop protection systems. Currently, innovative methods are being sought to support unit processes such as precision agrochemical application, while at the same time monitoring crop health and reducing herbicide usage. The development of innovative technology using convolutional neural networks (CNNs) for the automatic detection, classification, and localization of plant seedlings seems highly desirable. This will translate into obtaining localized spot-spraying capabilities based on the specific morphological structure and geometric positioning of the young plants. This study aims to apply convolutional neural networks to the classification and precise localization of radish, parsley, and cress seedlings supported by comparative image processing using the LAB and HSV color spaces combined with Otsu thresholding. The results show that LAB-based segmentation using the a channel is highly effective and stable when identifying plant morphological structures using CNNs, particularly under challenging illumination conditions. The use of CNNs, contour analysis, and geometric center (centroid) calculations together with dataset expansion via Auxiliary Classifier Generative Adversarial Networks (AC-GAN) shows the most favorable results. The best results were obtained for a sample during the intermediate growth stage (5–12 days) under sunny conditions, where the LAB-based approach achieved a precision of 0.89, recall of 0.84, and accuracy of 0.93. The CNN model achieved an overall classification accuracy of 93% on a balanced dataset comprising over 1,200 samples generated from approximately 60 original field images using AC-GAN, confirming its effectiveness for the classification of seedling microstructures under realistic field conditions.