17944398. WHITE-BOX TEMPERATURE SCALING FOR UNCERTAINTY ESTIMATION IN OBJECT DETECTION simplified abstract (Ford Global Technologies, LLC)
Contents
- 1 WHITE-BOX TEMPERATURE SCALING FOR UNCERTAINTY ESTIMATION IN OBJECT DETECTION
- 1.1 Organization Name
- 1.2 Inventor(s)
- 1.3 WHITE-BOX TEMPERATURE SCALING FOR UNCERTAINTY ESTIMATION IN OBJECT DETECTION - A simplified explanation of the abstract
- 1.4 Simplified Explanation
- 1.5 Potential Applications
- 1.6 Problems Solved
- 1.7 Benefits
- 1.8 Potential Commercial Applications
- 1.9 Possible Prior Art
- 1.10 Original Abstract Submitted
WHITE-BOX TEMPERATURE SCALING FOR UNCERTAINTY ESTIMATION IN OBJECT DETECTION
Organization Name
Inventor(s)
Sandhya Bhaskar of Sunnyvale CA (US)
Nikita Jaipuria of Pittsburgh PA (US)
Jinesh Jain of San Francisco CA (US)
Shreyasha Paudel of Sunnyvale CA (US)
WHITE-BOX TEMPERATURE SCALING FOR UNCERTAINTY ESTIMATION IN OBJECT DETECTION - A simplified explanation of the abstract
This abstract first appeared for US patent application 17944398 titled 'WHITE-BOX TEMPERATURE SCALING FOR UNCERTAINTY ESTIMATION IN OBJECT DETECTION
Simplified Explanation
The patent application describes a system and method for determining uncertainty estimation in an object detection deep neural network (DNN) by utilizing a calibration dataset to correct for class imbalance and update output data sets of the DNN.
- Retrieval of a calibration dataset from a validation dataset
- Determination of background ground truth boxes by comparing ground truth boxes with detection boxes using an IoU threshold
- Correction for class imbalance by updating the ground truth class to include background ground truth boxes
- Estimation of uncertainty based on the class imbalance correction
- Updating output data sets based on the class imbalance correction
Potential Applications
This technology could be applied in various fields such as autonomous driving, surveillance systems, and medical imaging for more accurate object detection and uncertainty estimation.
Problems Solved
This technology addresses the issue of class imbalance in object detection DNNs, which can lead to inaccurate predictions and uncertainty estimation.
Benefits
The system and method described in the patent application improve the reliability and accuracy of object detection DNNs by correcting for class imbalance and providing more precise uncertainty estimation.
Potential Commercial Applications
Potential commercial applications of this technology include enhancing the performance of autonomous vehicles, improving security systems, and enhancing medical diagnostic tools.
Possible Prior Art
One possible prior art could be methods for uncertainty estimation in deep learning models, but specific techniques for correcting class imbalance in object detection DNNs may be novel.
What are the specific techniques used to correct for class imbalance in the object detection DNN?
The specific techniques involve updating the ground truth class to include background ground truth boxes based on the number of ground truth boxes in the ground truth class, thus correcting for class imbalance.
How does the system estimate uncertainty based on the class imbalance correction?
The system estimates uncertainty by taking into account the corrected class imbalance, which provides a more accurate representation of the object detection DNN's performance and reliability.
Original Abstract Submitted
A system and method includes determining uncertainty estimation in an object detection deep neural network (DNN) by retrieving a calibration dataset from a validation dataset that includes scores associated with all classes in an image, including a background (BG) class, determining background ground truth boxes in the calibration dataset by comparing ground truth boxes with detection boxes generated by the object detection DNN using an intersection over union (IoU) threshold, correcting for class imbalance between ground truth boxes and background ground truth boxes in a ground truth class by updating the ground truth class to include a number of background ground truth boxes based on a number of ground truth boxes in the ground truth class, estimating uncertainty of the object detection DNN based on the class imbalance correction, and updating output data sets of the object detection DNN based on the class imbalance correction.
