Demo
MDE using LR²Depth on low power mobile device
Monocular depth estimation (MDE) is crucial for various computer vision applications, but existing methods often struggle to balance inference speed and accuracy when processing large-region visual information. This paper introduces LR²Depth, a novel MDE method that addresses this challenge by utilizing large-kernel convolution on low-resolution feature maps for efficient large-region feature aggregation. Our approach leverages the fact that each pixel on low-resolution feature maps corresponds to a larger region of the original image, allowing for fast and accurate depth predictions at a lower inference cost. Extensive experiments on NYU-Depth-V2, KITTI, and SUN RGB-D datasets demonstrate that LR$^2$Depth not only achieves state-of-the-art performance but also operates approximately twice as fast as previous MDE methods. Notably, at the time of submission, LR²Depth secured the top-1 position on the KITTI depth prediction online benchmark in 2024.
Overview of LR²Depth. Left: The workflow and fundamental unit of LR²Depth. LR²Depth employs large convolution kernels exclusively at low resolution stages, aggregating information over large region with less budgets. Right: The impact of identically sized convolutional kernels on feature maps of different resolutions. It is evident that in low-resolution feature map, the kernel can easily traverse object to extract distance information.
