Generating Image Descriptions via Sequential Cross-Modal Alignment Guided by Human Gaze

30317

Knowledge Triples