Depth Map Generation Using Neural Network Bottleneck Circuits

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional autonomous vehicles rely on expensive LiDAR sensors for accurate distance measurement, hindering their widespread adoption due to the high cost of these sensors.

Innovation Solution

A device comprising an encoder, a bottleneck circuit, and a decoder that generates a depth map from camera images without using LiDAR sensors, utilizing neural networks and attention circuits to extract and process feature data, enabling distance information generation from RGB camera images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If LiDAR sensors are used for distance measurement, then measurement precision is improved, but device cost increases

Engineering Contradiction:

Improvedistance measurement accuracyVSAvoiddevice cost

Core Design Contradiction:

Measurement precisionVSEase of manufacture

Solution Approach 1:

The patent uses a camera to capture optical images as a copy or alternative representation of the scene, replacing the need for direct LiDAR distance measurement. The depth map generated from the image serves as a computational copy of distance information, achieving accurate depth estimation without expensive LiDAR hardware

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical LiDAR sensing system with a computational imaging approach using a camera and neural network processing. The encoder-decoder network with bottleneck circuits substitutes the physical distance measurement mechanism with an algorithmic depth map generation process, reducing hardware cost while maintaining measurement precision

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If LiDAR sensors are used for autonomous vehicle navigation, then reliability is improved, but device complexity increases

Engineering Contradiction:

Improveautonomous navigation reliabilityVSAvoidsensor system complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent makes the camera serve multiple functions: it captures both color information for object recognition and depth information for distance measurement. The same image input is processed to generate both semantic understanding and depth maps, eliminating the need for separate LiDAR hardware and reducing overall system complexity while maintaining navigation reliability

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent introduces a depth map as an intermediary representation that bridges the gap between 2D camera images and 3D spatial understanding. This intermediate depth information serves as a computational mediator that provides reliable distance data without requiring complex LiDAR sensor systems

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11481912B2Device for generating a depth map

Publication Date: 2022.10.25 SK HYNIX INC

AI summary

A device includes an encoder configured to generate a plurality of feature data by encoding an image; a bottleneck circuit configured to generate enhanced feature data from first bottleneck data among the plurality of feature data; and a decoder configured to generate a depth map corresponding to the image by decoding the enhanced feature data and the plurality of feature data except the first bottleneck data.