AI-Based Image Colorization and Object Recognition

Black-and-white photographs lack the color information that makes an image easy to interpret, and adding that color by hand is slow and subjective. This paper describes a web application that brings automatic image colorization and object recognition together in one workflow. The user uploads a grayscale image through a Streamlit interface, the image is preprocessed, and a pretrained DDColor model produces a realistic colorized version. If the user asks for it, the colorized image is then passed to a pretrained YOLOv8x detector, which returns the names of the objects it finds along with confidence scores and bounding boxes. The system is written in Python with PyTorch and uses CUDA for GPU acceleration, with an automatic fallback to the CPU. Both models are used in their pretrained form, so no custom training is needed. On a sample photograph, the system produced plausible colors, kept the structure of the original scene, and correctly identified the main objects in the colorized result. The work is meant for restoring historical photographs and for educational and documentation use. Index Terms— image colorization, object recognition, deep learning, DDColor, YOLOv8, Streamlit