The study investigates how the CLIP model can be applied in artwork classification and retrieval tasks. The researchers used the NoisyArt dataset, a collection of artwork images from the web, to perform exhaustive experiments. The study found that the CLIP model achieved impressive results in zero-shot classification and promising results in both artwork-to-artwork and description-to-artwork domains. This research highlights the potential of multimodal approaches, like CLIP, in improving the performance of tasks related to visual information.
Publication date: 21 Sep 2023
Project Page: https://arxiv.org/abs/2309.12110
Paper: https://arxiv.org/pdf/2309.12110
