Ad Creative Image Classification Tool for Marketing Teams

Marketing teams accumulate thousands of creative assets across campaigns, and finding every banner from last spring or separating social creatives from print scans becomes a manual slog nobody budgets for. Automatic classification of ad images is a small capability with outsized workflow value for agencies and in-house teams alike. The intern builds a Python tool where marketers upload ad images and a machine learning classifier labels each one by type, whether banner, social post, or print, along with a confidence score. OpenCV handles the preprocessing: resizing, color normalization, and feature extraction that make dimensions, aspect ratios, and layout patterns learnable. The intern curates a labeled training set, trains and evaluates the classifier with standard data science practice, and builds a simple upload-and-review flow where users can correct mislabeled predictions, with those corrections collected to improve future retraining runs. The project is a clean, end-to-end introduction to applied computer vision: dataset building, image preprocessing, model training and honest evaluation, and packaging the result as a tool a non-technical marketing team could genuinely fold into their daily work.