18189679. INSTANCE-AWARE REPEAT FACTOR SAMPLING METHOD FOR IMBALANCED DATA IN OBJECT DETECTION simplified abstract (Robert Bosch GmbH)
Contents
INSTANCE-AWARE REPEAT FACTOR SAMPLING METHOD FOR IMBALANCED DATA IN OBJECT DETECTION
Organization Name
Inventor(s)
BURHANEDDIN Yaman of San Jose CA (US)
CHUN HAO Liu of Redwood City CA (US)
INSTANCE-AWARE REPEAT FACTOR SAMPLING METHOD FOR IMBALANCED DATA IN OBJECT DETECTION - A simplified explanation of the abstract
This abstract first appeared for US patent application 18189679 titled 'INSTANCE-AWARE REPEAT FACTOR SAMPLING METHOD FOR IMBALANCED DATA IN OBJECT DETECTION
- Simplified Explanation:**
This patent application introduces an approach to handle long-tail distribution with data imbalance in image analysis, specifically object detection and classification, by enhancing Repeat Factor Sampling methods using both images and bounding boxes.
- Key Features and Innovation:**
- Improves Repeat Factor Sampling methods for training data generation.
- Considers both images and bounding boxes to address data imbalance.
- Applicable in image analysis domains like object detection and classification.
- Potential Applications:**
The technology can be utilized in various image analysis tasks, particularly in improving object detection and classification accuracy in scenarios with long-tail distribution and data imbalance.
- Problems Solved:**
- Addresses data imbalance in long-tail distribution scenarios.
- Enhances training data quality for improved performance in image analysis tasks.
- Benefits:**
- Improved accuracy in object detection and classification.
- Enhanced training data quality.
- Better handling of data imbalance in long-tail distribution scenarios.
- Commercial Applications:**
Enhancing object detection and classification accuracy can benefit industries such as surveillance, autonomous vehicles, and retail for improved decision-making processes and operational efficiency.
- Questions about the Technology:**
1. How does this technology improve training data quality in image analysis tasks? 2. What are the specific challenges faced in handling data imbalance in long-tail distribution scenarios?
Original Abstract Submitted
An approach for handling long-tail distribution with data imbalance. Disclosed embodiments improve Repeat Factor Sampling methods by considering both images and bounding boxes (i.e., instances) to generate improved sets of training data. Disclosed embodiments may be useful in image analysis domains, such as object detection and classification.
