18518609. IMAGE ANALYSIS METHOD AND IMAGE ANALYSIS SYSTEM simplified abstract (MEDIATEK Inc.)

IMAGE ANALYSIS METHOD AND IMAGE ANALYSIS SYSTEM

Organization Name

MEDIATEK Inc.

Inventor(s)

Ting-Hsuan Liao of Hsinchu City (TW)

Huang-Ru Liao of Hsinchu City (TW)

Shan-Ya Yang of Hsinchu City (TW)

Jie-En Yao of Hsinchu City (TW)

Li-Yuan Tsao of Hsinchu City (TW)

Hsu-Shen Liu of Hsinchu City (TW)

Bo-Wun Cheng of Hsinchu City (TW)

Chen-Hao Chao of Hsinchu City (TW)

Chia-Che Chang of Hsinchu City (TW)

Yi-Chen Lo of Hsinchu City (TW)

Chun-Yi Lee of Hsinchu City (TW)

IMAGE ANALYSIS METHOD AND IMAGE ANALYSIS SYSTEM - A simplified explanation of the abstract

This abstract first appeared for US patent application 18518609 titled 'IMAGE ANALYSIS METHOD AND IMAGE ANALYSIS SYSTEM

Simplified Explanation

The Edge Learning based Domain Adaptation (ELDA) framework incorporates edge information into its training process to serve as domain invariant information, outperforming contemporary methods on semantic segmentation based UDA tasks.

• ELDA framework incorporates edge information into training process
• Demonstrated to outperform state-of-the-art methods on semantic segmentation based UDA tasks
• Better separates feature distributions of different classes

Potential Applications

The ELDA framework can be applied in various fields such as computer vision, image processing, and artificial intelligence for improving domain adaptation tasks.

Problems Solved

ELDA addresses the challenges of high extraction costs and unreliable prediction quality associated with using depth as domain invariant information in UDA tasks.

Benefits

- Improved performance in semantic segmentation based UDA tasks - Better separation of feature distributions of different classes - Reduced extraction costs and improved prediction quality

Potential Commercial Applications

The ELDA framework can be utilized in industries such as autonomous driving, medical imaging, and surveillance systems for enhancing the accuracy and efficiency of domain adaptation tasks.

Possible Prior Art

Prior art in domain adaptation methods using depth information as domain invariant information can be found in various research papers and patents related to computer vision and machine learning.

How does ELDA compare to other edge-based domain adaptation methods in terms of performance and efficiency?

Further comparative studies with other edge-based domain adaptation methods are needed to fully understand the advantages and limitations of ELDA in different scenarios.

What are the potential limitations or challenges of implementing the ELDA framework in real-world applications?

Exploring the scalability, robustness, and adaptability of ELDA in diverse real-world scenarios is essential to identify any potential limitations or challenges in its practical implementation.

Original Abstract Submitted

Many unsupervised domain adaptation (UDA) methods have been proposed to bridge the domain gap by utilizing domain invariant information. Most approaches have chosen depth as such information and achieved remarkable successes. Despite their effectiveness, using depth as domain invariant information in UDA tasks may lead to multiple issues, such as excessively high extraction costs and difficulties in achieving a reliable prediction quality. As a result, we introduce Edge Learning based Domain Adaptation (ELDA), a framework which incorporates edge information into its training process to serve as a type of domain invariant information. Our experiments quantitatively and qualitatively demonstrate that the incorporation of edge information is indeed beneficial and effective, and enables ELDA to outperform the contemporary state-of-the-art methods on two commonly adopted benchmarks for semantic segmentation based UDA tasks. In addition, we show that ELDA is able to better separate the feature distributions of different classes.