Cross-modal Transfer Between Vision and Language for Protest Detection
Explore a conference talk on cross-modal transfer between vision and language for protest detection. Delve into the potential of multimodal data in socio-political event detection, focusing on a proposed approach that leverages existing annotated unimodal data for zero-shot event detection across different modalities. Learn about the application of a pretrained vision-and-language alignment model (CLIP) for protest detection in text and images. Discover how annotated protest text data can supplement protest detection in images, and vice versa. Gain insights from speakers Ria Raj and Kajsa Andreasson, who share their expertise in machine learning, natural language processing, and artificial intelligence techniques applied to threat intelligence and text analytics.