Using Language to Extend to Unseen Domains
This addresses the challenge of domain adaptation for vision models, reducing the need for costly data collection in new domains.
The paper tackles the problem of expensive data collection for vision models across multiple domains by using language descriptions to extend models to unseen domains without additional images. The method LADS outperforms standard fine-tuning and ensemble approaches on four domain adaptation benchmarks.
It is expensive to collect training data for every possible domain that a vision model may encounter when deployed. We instead consider how simply verbalizing the training domain (e.g. "photos of birds") as well as domains we want to extend to but do not have data for (e.g. "paintings of birds") can improve robustness. Using a multimodal model with a joint image and language embedding space, our method LADS learns a transformation of the image embeddings from the training domain to each unseen test domain, while preserving task relevant information. Without using any images from the unseen test domain, we show that over the extended domain containing both training and unseen test domains, LADS outperforms standard fine-tuning and ensemble approaches over a suite of four benchmarks targeting domain adaptation and dataset bias.