I Can't Believe There's No Images! Learning Visual Tasks Using only Language Supervision
This addresses the challenge of reducing reliance on costly visual data for computer vision tasks, offering a novel approach for domains with limited image resources, though it builds incrementally on contrastive embedding methods.
The paper tackles the problem of learning visual tasks using only language supervision, without training on visual data, and finds that models trained solely on text perform close to image-trained models on tasks like image captioning and visual entailment, surpassing prior work by over 9 points in some cases and by over 30 points on visual news captioning.
Many high-level skills that are required for computer vision tasks, such as parsing questions, comparing and contrasting semantics, and writing descriptions, are also required in other domains such as natural language processing. In this paper, we ask whether it is possible to learn those skills from text data and then transfer them to vision tasks without ever training on visual training data. Key to our approach is exploiting the joint embedding space of contrastively trained vision and language encoders. In practice, there can be systematic differences between embedding spaces for different modalities in contrastive models, and we analyze how these differences affect our approach and study strategies to mitigate this concern. We produce models using only text training data on four representative tasks: image captioning, visual entailment, visual question answering and visual news captioning, and evaluate them on standard benchmarks using images. We find these models perform close to models trained on images, while surpassing prior work for captioning and visual entailment in this text-only setting by over 9 points, and outperforming all prior work on visual news by over 30 points. We also showcase a variety of stylistic image captioning models that are trained using no image data and no human-curated language data, but instead using readily-available text data from books, the web, or language models.