Weed in farming is a serious issue that affects crop productivity by competing with the crops for necessary resources like nutrients, water, and sunlight. Weeds may also harbor pests and diseases, thus further affecting the growth of crops. Early Detection of weeds is very important in preventing the adverse effects. Conventional machine learning methods of differentiating weeds from crops have not been effective and reliable, especially in the initial growth stage. This work investigates plant seedling classification using deep learning. A new Convolutional Neural Network (CNN) architecture is proposed for classifying seedlings during their early growth stage. The performance of the model is tested with a plant seedling dataset and evaluated based on accuracy, precision, recall, and F1 score. Experimental outcomes show that the system can accurately distinguish 12 species, such as 3 crops and 9 weeds, with the average classification accuracy of 91.37%. Comparison with current plant seedling classification techniques reflects the advantages of the proposed method.