# CNN-based synthesis of realistic high-resolution LiDAR data

## TL;DR

This paper introduces a CNN-based method for synthesizing high-resolution LiDAR data that produces realistic, semantically meaningful point clouds with improved geometric accuracy, validated through extensive experiments.

## Contribution

The paper proposes a novel CNN approach with specialized loss functions for realistic high-res LiDAR data synthesis, addressing missing data and aligning with real sensor quality.

## Key findings

- Significant improvement in geometric accuracy over traditional methods
- Enhanced semantic segmentation performance using generated data
- High perceptual quality confirmed by mean opinion scores

## Abstract

This paper presents a novel CNN-based approach for synthesizing high-resolution LiDAR point cloud data. Our approach generates semantically and perceptually realistic results with guidance from specialized loss-functions. First, we utilize a modified per-point loss that addresses missing LiDAR point measurements. Second, we align the quality of our generated output with real-world sensor data by applying a perceptual loss. In large-scale experiments on real-world datasets, we evaluate both the geometric accuracy and semantic segmentation performance using our generated data vs. ground truth. In a mean opinion score testing we further assess the perceptual quality of our generated point clouds. Our results demonstrate a significant quantitative and qualitative improvement in both geometry and semantics over traditional non CNN-based up-sampling methods.