Multi-task CNN for ADAS Scene Understanding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current scene understanding systems and advanced driver-assistance systems require multiple computing resources and ignore shared features between perception tasks like object detection and semantic segmentation, leading to inefficiencies.
Innovation Solution
A multi-task Convolutional Neural Network (CNN) is used to extract shared features across different perception tasks, allowing concurrent processing of these tasks in a single pass to provide multiple perception task outputs, which are then used to form a parametric representation of a driving scene for collision avoidance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple independent CNNs are used for different perception tasks, then each task can be processed independently, but computing resources and device complexity increase significantly
Solution Approach 1:
The patent merges multiple independent CNNs into a single multi-task CNN that shares common feature extraction layers. Different perception tasks (object detection, semantic segmentation, depth estimation) are processed concurrently within one network architecture, reducing the number of separate computing units required while maintaining task independence through separate task-specific output layers.
Solution Approach 2:
The multi-task CNN is designed with universal feature extraction capabilities that serve multiple perception tasks simultaneously. The shared backbone network extracts general features that can be utilized by different task-specific heads, making the system multi-functional without requiring separate specialized networks for each task.
2Reliability
If multiple independent CNNs are used for different perception tasks, then each task has dedicated processing, but the system ignores mutual features between tasks
Solution Approach 1:
By merging multiple CNNs into one multi-task network, the patent enables feature sharing across different perception tasks. The shared layers extract mutual features that are beneficial to multiple tasks, preventing information loss while still allowing task-specific processing through separate output layers that receive the shared features.
Data Source
AI summary
A method is provided in an Advanced Driver-Assistance System (ADAS). The method extracts, from an input video stream including a plurality of images using a multi-task Convolutional Neural Network (CNN), shared features across different perception tasks. The perception tasks include object detection and other perception tasks. The method concurrently solves, using the multi-task CNN, the different perception tasks in a single pass by concurrently processing corresponding ones of the shared features by respective different branches of the multi-task CNN to provide a plurality of different perception task outputs. Each respective different branch corresponds to a respective one of the different perception tasks. The method forms a parametric representation of a driving scene as at least one top-view map responsive to the plurality of different perception task outputs. The method controls an operation of the vehicle for collision avoidance responsive to the at least one top-view map indicating an impending collision.