Compressed Log Depth Map Generation for AI Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital image processing systems face challenges in accuracy, efficiency, and flexibility when estimating depth maps due to their inability to accurately differentiate objects at varying distances and retain important depth structure information, leading to inefficiencies in training AI models and downstream tasks.
Innovation Solution
The implementation of machine learning models utilizing compressed log scene measurement maps, which convert depth maps to disparity maps and apply a logarithmic function based on distance distribution metrics to generate accurate and flexible depth estimates, improving the separation of objects at different depths and reducing computational resources needed for training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional depth estimation systems use artificial intelligence models to generate depth maps, then depth estimation can be performed, but the accuracy in differentiating objects at varying distances deteriorates
Solution Approach 1:
The patent transforms the depth map representation by applying a logarithmic function to convert depth values into a compressed log depth map. This parameter transformation changes the scale and distribution of depth values, allowing for better differentiation of objects at varying distances while preserving depth structure information that was lost in conventional depth estimation
Solution Approach 2:
The patent introduces compressed log depth maps as an intermediary representation between the original depth map and the final depth estimation output. This intermediate form serves as a mediator that retains important depth structure information while improving the accuracy of depth differentiation for objects at various distances
2Productivity
If conventional systems use standard depth maps for downstream image manipulation tasks, then processing can be performed, but the efficiency and quality of tasks such as background blurring deteriorate
Solution Approach 1:
The patent performs preliminary transformation of the depth map into a compressed log depth map before conducting downstream image manipulation tasks. This pre-processing step prepares the depth information in an optimized format that improves both the efficiency and quality of subsequent operations such as background blurring, edge detection, and scene segmentation
3Measurement precision
If machine learning models are trained with standard depth maps, then training can be performed, but the training time and computational resources increase
Solution Approach 1:
The patent applies parameter transformation by converting standard depth maps to compressed log depth maps for training data preparation. This transformation changes the distribution and scale of depth values, enabling machine learning models to learn more efficiently from the transformed data, thereby reducing training time and computational resource requirements while maintaining or improving depth map accuracy
Data Source
AI summary
The present disclosure relates to systems, non-transitory computer-readable media, and methods for training and/or implementing machine learning models utilizing compressed log scene measurement maps. For example, the disclosed system generates compressed log scene measurement maps by converting scene measurement maps to compressed log scene measurement maps by applying a logarithmic function. In particular, the disclosed system uses scene measurement distribution metrics from a digital image to determine a base for the logarithmic function. In this way, the compressed log scene measurement maps normalize ranges within a digital image and accurately differentiates between scene elements objects at a variety of depths. Moreover, for training, the disclosed system generates a predicted scene measurement map via a machine learning model and compares the predicted scene measurement map with a compressed log ground truth map. By doing so, the disclosed system trains the machine learning model to generate accurate compressed log depth maps.