Biome Explorer App

Deep Learning & Entrepreneurship module at BUas, Year 1 Block C. Individual project combining deep learning with human-centered design.

Deep learning image classifier that identifies 7 global biomes from photos. Built through 4 iterative model experiments culminating in a VGG16 transfer-learning model (~90% test accuracy), with GradCAM + LIME interpretability analysis and a user-tested Figma mobile app prototype.

Key Highlights

01.Trained 4 model iterations progressing from a baseline CNN to VGG16 transfer learning, achieving ~90% test accuracy.
02.Applied GradCAM and LIME to explain model decisions and analyse which visual features drive biome classification.
03.Conducted a Think-aloud + A/B user study to validate two UI variants; iterated to produce a final high-fidelity Figma wireframe.