CoHAtNet: An integrated convolutional-transformer architecture with hybrid self-attention for end-to-end camera localization | Publicación