Can I read X-Linear Attention for Image Captioning on EtoBox?
X-Linear Attention for Image Captioning by 23bq5a1204 is a document available to read on EtoBox.
What is X-Linear Attention for Image Captioning about?
The document introduces X-Linear Attention Networks (X-LAN) for image captioning, which utilizes a novel X-Linear attention block to capture higher order interactions between visual and textual data. This approach enhances multi-modal reasoning by employing bilinear pooling to model 2nd order interactions, significantly improving performance on the COCO dataset. Experiments show that X-LAN achieves state-of-the-art results, with a CIDEr score of 132.8% when integrated with Transformer mechanisms.
- Author
- 23bq5a1204
- Language
- EN