LightDepth: A Resource Efficient Depth Estimation Approach for Dealing with Ground Truth Sparsity via Curriculum Learning

TL;DR

LightDepth introduces a resource-efficient, curriculum-based depth estimation method that maintains high accuracy while significantly reducing response time, suitable for autonomous devices with limited resources.

Contribution

The paper proposes a novel, model-agnostic curriculum learning approach for depth estimation that enhances efficiency without sacrificing accuracy.

Findings

Achieves state-of-the-art accuracy in depth estimation.

Reduces response time by 71% compared to existing models.

Provides a resource-efficient solution suitable for autonomous devices.

Abstract

Advances in neural networks enable tackling complex computer vision tasks such as depth estimation of outdoor scenes at unprecedented accuracy. Promising research has been done on depth estimation. However, current efforts are computationally resource-intensive and do not consider the resource constraints of autonomous devices, such as robots and drones. In this work, we present a fast and battery-efficient approach for depth estimation. Our approach devises model-agnostic curriculum-based learning for depth estimation. Our experiments show that the accuracy of our model performs on par with the state-of-the-art models, while its response time outperforms other models by 71%. All codes are available online at https://github.com/fatemehkarimii/LightDepth.