Packaging design has become a central element of consumer choice and brand recognition, which are now more affected by artificial intelligence (AI) and new media technology. With increasing demands in complexity and personalization of design, intelligent systems provide new approaches to automate, create and optimize visual packaging. The proposed study will develop a multi-model AI system to improve advertisement packaging via classification, creative system generation, object detection, and optimization of design through improved deep learning methods. The framework combines four approaches: the Convolutional Neural Networks (CNNs) that classify the visual elements of design, Generative Adversarial Networks (GANs) that create the new packaging designs, YOLO that can detect objects in real-time, and Reinforcement Learning (RL) that can optimize the package design in terms of color and logo size. Normalization, PCA and clustering were applied to preprocess ADS16 and additional designs. The models were assessed based on accuracy, F1-score, FID, IS, mAP, and reward trajectory in order to measure functional and aesthetic performance. The suggested multi-model is an effective combination of AI and design intelligence. Specifically, reinforcement learning offers the most promising prospects of developing customized, data-reinforced packaging solutions in accordance with the new media and advertising objectives.
