AI-Driven Generative 3D Metric-Scale Monocular SLAM. Novel Dual Convolution Vision Transformer DCViT 

KITTI 10
Loop closure is disabled to demonstrate the effectiveness of DCViT's metric scale depth prediction.
A novel convolutional vision transformer deep learning architecture is introduced to generate metric scale 3D depth predictions from monocular images. All information can be found in my PhD thesis. The source code will be released soon.