Viral Images: Identifying Reprintings within 1.5 Million Photographs in Chronicling America
Researchers have developed a new method to identify reprinted images within the Chronicling America historical newspaper collection. The project, named Viral Images, utilizes contrastive language-image pretraining (CLIP) to embed and cluster 1.5 million photographs. This approach allows for the discovery of visual content that circulated across different newspapers over time, with a public interface enabling interactive study of these identified clusters. AI