Using advanced computer vision and machine learning algorithms, an artificial intelligence-based weed detection system automatically identifies and classifies weeds in agricultural fields. The system employs hardware components such as cameras or drones to capture high-resolution images or videos of the fields. These images are analyzed by AI algorithms trained on extensive weed image datasets to distinguish weeds from crops accurately. The AI processes images by extracting features like leaf shape, color, texture, and growth patterns to identify weeds. It provides real-time feedback, enabling farmers to take timely actions for weed control. This study explores the application of AI in weed detection, presenting an effective solution for automating this task. We compared the performance of the SVM classifier with a CNN classifier and found that the CNN alone is insufficient. Consequently, we used an ensemble model, which combined a weighted CNN (60%) and a Random Forest (40%), outperforming the CNN model. The goal is to develop a system that sprays pesticides only on weeds, minimizing pesticide use and waste. For validation, we used a dataset of 1300 labeled images of sesame crops and various weeds, formatted in YOLO, and sized 512 × 512 pixels. Results showed that the SVM achieved 96.8% accuracy, a custom 5-layer CNN achieved 97.7% accuracy, and our ensemble model achieved the highest validation accuracy of 99.725%.