Advanced zero-shot image classification solution for CVPR 2025 challenge. Leverages CLIP architecture with novel prompt engineering, clustering analysis, and ensemble methods to classify unseen categories without training data, achieving state-of-the-art performance.
embeddings artificial-neural-networks convolutional-neural-networks vizwiz-dataset image-net-dataset cvpr-challenge blind-images-training-model
-
Updated
Jun 8, 2025 - Jupyter Notebook