AI Colour Detection System – Deep Learning Image

About the project

Developed a deep-learning image classification system designed to identify colours from real-world photographs while accounting for variations caused by lighting, white balance, and image compression. Instead of relying on raw pixel values, the system uses a dedicated preprocessing and normalization pipeline before passing image data to the neural network.The project focused heavily on improving input preprocessing to reduce the impact of inconsistent lighting conditions and provide the classifier with more standardized image information. Neural network models were then trained and evaluated to classify colours from the processed images.Experimentation demonstrated that improvements to preprocessing and normalization had a greater impact on classification performance than simply increasing model complexity, highlighting the importance of data preparation in practical computer-vision systems.