View source for 17957689. CNN SEAMLESS TILE PROCESSING FOR LOW-POWER INFERENCE ACCELERATOR simplified abstract (ATI TECHNOLOGIES ULC)

=CNN SEAMLESS TILE PROCESSING FOR LOW-POWER INFERENCE ACCELERATOR=


==Organization Name==

[[:Category:ATI TECHNOLOGIES ULC|ATI TECHNOLOGIES ULC]]


[[Category:ATI TECHNOLOGIES ULC]]

==Inventor(s)==

[[:Category:Tung Chuen Kwong of Markham (CA)|Tung Chuen Kwong of Markham (CA)]][[Category:Tung Chuen Kwong of Markham (CA)]]

[[:Category:Ying Liu of Markham (CA)|Ying Liu of Markham (CA)]][[Category:Ying Liu of Markham (CA)]]

[[:Category:Akila Subramaniam of Austin TX (US)|Akila Subramaniam of Austin TX (US)]][[Category:Akila Subramaniam of Austin TX (US)]]

==CNN SEAMLESS TILE PROCESSING FOR LOW-POWER INFERENCE ACCELERATOR - A simplified explanation of the abstract==


This abstract first appeared for US patent application 17957689 titled 'CNN SEAMLESS TILE PROCESSING FOR LOW-POWER INFERENCE ACCELERATOR

==Simplified Explanation==

The abstract describes a patent application for a method and device that processes image data using layers of a convolutional neural network on a sub-frame portion basis.

* The processor determines a receptive field for an input tile of an image through backward propagation.
* The size of the input tile is determined based on the receptive field and the amount of local memory allocated to store data for the input tile.
* The processor checks if the allocated local memory is sufficient to store the data of the input tile and padded data for the receptive field.

== Potential Applications ==
This technology could be applied in image recognition systems, autonomous vehicles, medical imaging, and surveillance systems.

== Problems Solved ==
This technology solves the problem of efficiently processing image data on a sub-frame portion basis using convolutional neural networks.

== Benefits ==
The benefits of this technology include improved image processing efficiency, reduced memory usage, and enhanced performance of convolutional neural networks.

== Potential Commercial Applications ==
The potential commercial applications of this technology include image processing software, hardware for image analysis, and systems for real-time image recognition.

== Possible Prior Art ==
One possible prior art for this technology could be the use of convolutional neural networks for image processing in various fields such as computer vision and pattern recognition.

=== What are the specific technical details of the backward propagation process mentioned in the abstract? ===

The specific technical details of the backward propagation process involve calculating the gradients of the loss function with respect to the weights of the neural network. This process helps in updating the weights of the network to minimize the loss function during training.

=== How does the determination of the receptive field impact the overall performance of the convolutional neural network? ===

The determination of the receptive field helps in capturing spatial dependencies in the input data, which is crucial for tasks like image recognition. A proper receptive field size ensures that the network can effectively learn features at different scales, leading to improved performance in tasks such as object detection and segmentation.


==Original Abstract Submitted==

Methods and devices are provided for processing image data on a sub-frame portion basis using layers of a convolutional neural network. The processing device comprises memory and a processor. The processor is configured to determine, for an input tile of an image, a receptive field via backward propagation and determine a size of the input tile based on the receptive field and an amount of local memory allocated to store data for the input tile. The processor determines whether the amount of local memory allocated to store the data of the input tile and padded data for the receptive field.



[[Category:G06T1/60]]