# SDNet: Semantically Guided Depth Estimation Network

## TL;DR

SDNet is a novel neural network that jointly predicts depth and semantic labels from a single image, improving accuracy and efficiency by leveraging shared features and ordinal classification.

## Contribution

The paper introduces SDNet, a unified model for simultaneous depth and semantic prediction, demonstrating enhanced performance and computational efficiency over separate models.

## Key findings

- Achieves state-of-the-art results in depth estimation and semantic segmentation.
- Joint prediction improves accuracy and reduces computational costs.
- Utilizes ordinal classification for depth estimation.

## Abstract

Autonomous vehicles and robots require a full scene understanding of the environment to interact with it. Such a perception typically incorporates pixel-wise knowledge of the depths and semantic labels for each image from a video sensor. Recent learning-based methods estimate both types of information independently using two separate CNNs. In this paper, we propose a model that is able to predict both outputs simultaneously, which leads to improved results and even reduced computational costs compared to independent estimation of depth and semantics. We also empirically prove that the CNN is capable of learning more meaningful and semantically richer features. Furthermore, our SDNet estimates the depth based on ordinal classification. On the basis of these two enhancements, our proposed method achieves state-of-the-art results in semantic segmentation and depth estimation from single monocular input images on two challenging datasets.