Samsung electronics co., ltd. (20240127589). HARDWARE FRIENDLY MULTI-KERNEL CONVOLUTION NETWORK simplified abstract
Contents
- 1 HARDWARE FRIENDLY MULTI-KERNEL CONVOLUTION NETWORK
- 1.1 Organization Name
- 1.2 Inventor(s)
- 1.3 HARDWARE FRIENDLY MULTI-KERNEL CONVOLUTION NETWORK - 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
- 1.9.1 Unanswered Questions
- 1.9.2 How does this technology compare to existing methods for processing feature maps in terms of computational efficiency and accuracy?
- 1.9.3 What impact could this technology have on the development of deep learning models and their applications in real-world scenarios?
- 1.10 Original Abstract Submitted
HARDWARE FRIENDLY MULTI-KERNEL CONVOLUTION NETWORK
Organization Name
Inventor(s)
Qingfeng Liu of San Diego CA (US)
Mostafa El-khamy of San Diego CA (US)
Sukhwan Lim of Los Altos Hills CA (US)
HARDWARE FRIENDLY MULTI-KERNEL CONVOLUTION NETWORK - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240127589 titled 'HARDWARE FRIENDLY MULTI-KERNEL CONVOLUTION NETWORK
Simplified Explanation
The abstract describes a system and method for processing and combining feature maps using a hardware-friendly multi-kernel convolution block (HFMCB). The method involves splitting an input feature map into multiple feature maps with reduced channels, processing each with different kernels, and then combining the processed feature maps.
- The system and method involve splitting an input feature map into multiple feature maps with reduced channels.
- Each of the feature maps is processed with a different series of kernels.
- The processed feature maps are then combined to produce the final output.
Potential Applications
This technology could be applied in image recognition systems, object detection algorithms, and video processing applications.
Problems Solved
This technology helps in reducing computational complexity, improving efficiency in processing feature maps, and enhancing the accuracy of convolutional neural networks.
Benefits
The benefits of this technology include faster processing speeds, reduced hardware requirements, improved accuracy in feature map processing, and enhanced performance in deep learning applications.
Potential Commercial Applications
The potential commercial applications of this technology include in the fields of autonomous vehicles, surveillance systems, medical imaging, and industrial automation.
Possible Prior Art
One possible prior art could be the use of traditional convolutional neural network architectures for processing feature maps, which may not be as efficient or hardware-friendly as the HFMCB method described in this patent application.
Unanswered Questions
How does this technology compare to existing methods for processing feature maps in terms of computational efficiency and accuracy?
This technology aims to improve computational efficiency and accuracy in processing feature maps by utilizing a hardware-friendly multi-kernel convolution block. However, a direct comparison with existing methods would provide a clearer understanding of its advantages.
What impact could this technology have on the development of deep learning models and their applications in real-world scenarios?
Understanding the potential impact of this technology on the development and deployment of deep learning models in real-world applications would shed light on its practical significance and potential for widespread adoption.
Original Abstract Submitted
a system and a method are disclosed for processing and combining feature maps using a hardware friendly multi-kernel convolution block (hfmcb). the method including splitting an input feature map into a plurality of feature maps, each of the plurality of feature maps having a reduced number of channels; processing each of the plurality of feature maps with a different series of kernels; and combining the processed plurality of feature maps.
