17546901. TRAINING SELF-CLASSIFIER USING IMAGE AUGMENTATIONS AND UNIFORM PRIOR simplified abstract (INTERNATIONAL BUSINESS MACHINES CORPORATION)
TRAINING SELF-CLASSIFIER USING IMAGE AUGMENTATIONS AND UNIFORM PRIOR
Organization Name
INTERNATIONAL BUSINESS MACHINES CORPORATION
Inventor(s)
Elad Amrani of Zikhron Yaakov (IL)
TRAINING SELF-CLASSIFIER USING IMAGE AUGMENTATIONS AND UNIFORM PRIOR - A simplified explanation of the abstract
This abstract first appeared for US patent application 17546901 titled 'TRAINING SELF-CLASSIFIER USING IMAGE AUGMENTATIONS AND UNIFORM PRIOR
Simplified Explanation
The abstract of the patent application describes a system that uses image samples for training a self-classifier neural network. The system includes a processor that can generate different augmented views of each image sample and then train the neural network using these augmented views. The goal is to minimize the cross entropy of the different augmented views by asserting a uniform prior on class predictions.
- The system uses image samples for training a self-classifier neural network.
- The processor generates different augmented views of each image sample.
- The neural network is trained using these augmented views.
- The goal is to minimize the cross entropy of the different augmented views.
- A uniform prior is asserted on class predictions.
Potential Applications
- Image recognition and classification systems
- Object detection and tracking systems
- Autonomous vehicles and robotics
- Medical imaging analysis
Problems Solved
- Improving the accuracy and performance of self-classifier neural networks
- Addressing the limitations of training neural networks with limited image samples
- Enhancing the ability to classify and recognize objects in various conditions and perspectives
Benefits
- Increased accuracy and robustness of image recognition and classification systems
- Improved performance and adaptability of object detection and tracking systems
- Enhanced capabilities of autonomous vehicles and robotics in understanding and interacting with the environment
- More accurate and efficient analysis of medical images for diagnosis and treatment planning
Original Abstract Submitted
An example system includes a processor to receive image samples for training. The processor can generate different augmented views of each of the image samples. The processor can then train a self-classifier neural network using the different augmented views to minimize a cross entropy of the different augmented views in which a uniform prior is asserted on class predictions.