Road Image Synthesis With 3D Object Placement Across Frames
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating training data for machine learning models to detect objects on roads require significant time and effort due to the manual identification of positions for adding computer-generated images, which is inefficient.
Innovation Solution
An image generation device that automatically designates object drawing areas in consecutive captured images using three-dimensional recognition and generates additional images of objects based on text input and random noise, reducing the need for manual intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual identification is used to specify the position where CG image is added to captured images, then the position can be accurately determined, but it takes significant time and effort to generate training data
Solution Approach 1:
The system uses the captured images themselves to automatically determine the positions for adding CG images. By detecting road areas and visible regions from the captured images, the system self-identifies appropriate positions without requiring manual intervention, thus resolving the contradiction between position accuracy and time consumption
Solution Approach 2:
The system pre-processes captured images to identify road areas and calculate visible regions before adding CG images. This preliminary action of automatically determining positions based on image analysis eliminates the need for manual position specification while maintaining accuracy
2Stability of the object's composition
If CG images are added to consecutive captured images using manual position specification, then consistent object positions can be maintained, but the process becomes labor-intensive and inefficient
Solution Approach 1:
The system automatically determines object positions in consecutive images by detecting road areas and calculating visible regions from each captured image. This self-service approach maintains position consistency through algorithmic determination rather than manual specification, significantly improving productivity
Solution Approach 2:
The system transitions from two-dimensional image coordinates to three-dimensional spatial understanding by calculating visible regions based on depth information and vehicle position. This dimensional transformation enables automatic consistent positioning of objects across consecutive images without manual intervention
Data Source
AI summary
An image generation device for generating image data representing consecutive images for training, includes a processor. The processor is configured to: acquire image data representing consecutive captured images in which a road is captured; three-dimensionally recognize an area of a road in each captured image; designate an initial object drawing area at an arbitrary position among recognized areas of a road, in an initial image; designate a subsequent object drawing area to a position where it corresponds to the initial object drawing area, in a subsequent image; and generate data of an image obtained by adding an additional image of an arbitrary object to the object drawing area designated in each image of the consecutive captured images.