Researchers at the Kwame Nkrumah University of Science and Technology (KNUST) have developed a smarter way for artificial intelligence (AI) to analyze high-resolution images without losing important details.

Traditional AI models shrink images to save computing power, but this often reduces accuracy. The new approach, called WaveNet, keeps images sharp while making the process more efficient.

WaveNet uses a technique called the wavelet packet transform (WPT) to break images into smaller, meaningful parts before feeding them into AI models.

It also includes a special feature, the wavelet-adaptive efficient channel attention (WAECA) module, which helps the AI focus on the most useful parts of an image.

Tests on popular AI models like ResNet-50 and MobileNetV2 showed that WaveNet improves accuracy while significantly reducing processing costs.

For example, when tested on the Caltech-256 dataset, a WaveNet-enhanced ResNet-50 achieved 72.47% accuracy, outperforming the standard version (70.65%) while using far fewer computational resources.

By making AI models both smarter and faster, this innovation could help improve image recognition in applications like medical imaging, surveillance, and autonomous vehicles, without requiring expensive hardware upgrades.

This research, published in Engineering Reports, was supported by the Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH through the Responsible Artificial Intelligence Lab at KNUST.

Authors of the studies include: Albert Dede, Henry Nunoo-Mensah, Emmanuel Kofi Akowuah, Kwame Osei Boateng, Prince Ebenezer Adjei, Francisca Adoma Acheampong, Isaac Acquah, Jerry John Kponyo.