GSAP: A Global Structure Attention Pooling Method for Graph-Based Visual Place Recognition

The Visual Place Recognition problem aims to use an image recognize the location that has been visited before. In most of scenes revisited, appearance and view are drastically different. Most previous works focus on 2-D image-based deep learning method. However, convolutional features not robust enough challenging mentioned above. this paper, in order take advantage information helps task these scenes, we propose a new graph construction approach extract useful from RGB depth fuse them data. Then, deal with as classification problem. We Global Pooling method—Global Structure Attention (GSAP), which improves accuracy by improving expression ability component. experiments show our GSAP method approximately 2–5%, 4–6%, whole model is change change.