Predicting Deep Zero-Shot Convolutional Neural Networks using Textual Descriptions
This work addresses the problem of zero-shot learning in computer vision by leveraging textual data to avoid manual attribute annotation, offering a novel approach for classifying unseen categories.
The paper tackles the challenge of zero-shot learning for visual categories by using textual descriptions to predict the output weights of convolutional and fully connected layers in a deep CNN, achieving significant performance improvements over previous methods on bird and flower datasets.
One of the main challenges in Zero-Shot Learning of visual categories is gathering semantic attributes to accompany images. Recent work has shown that learning from textual descriptions, such as Wikipedia articles, avoids the problem of having to explicitly define these attributes. We present a new model that can classify unseen categories from their textual description. Specifically, we use text features to predict the output weights of both the convolutional and the fully connected layers in a deep convolutional neural network (CNN). We take advantage of the architecture of CNNs and learn features at different layers, rather than just learning an embedding space for both modalities, as is common with existing approaches. The proposed model also allows us to automatically generate a list of pseudo- attributes for each visual category consisting of words from Wikipedia articles. We train our models end-to-end us- ing the Caltech-UCSD bird and flower datasets and evaluate both ROC and Precision-Recall curves. Our empirical results show that the proposed model significantly outperforms previous methods.