Deep-learning-based image processing method and system

Patent metadata

Field	Value
Publication number	US-11614964-B2
Application number	US-201917642371-A
Country	US
Kind code	B2
Filing date	Dec 30, 2019
Priority date	Sep 12, 2019
Publication date	Mar 28, 2023
Grant date	Mar 28, 2023

Abstract

An image processing method is provided, which is applied to a deep learning model. A cache queue is provided in front of each layer of the deep learning model; a plurality of computation tasks are preset for each layer of the deep learning model in advance, and are configured for computing weight parameters and corresponding to-be-processed data in a plurality of channels in each corresponding layer in parallel, and storing a computation result into a cache queue behind the corresponding layer thereof; in addition, as long as the cache queue in front of the layer includes the computation result stored in the previous layer, the layer can obtain the to-be-processed data from the computation result, subsequent computation is performed, and a parallel pipeline computation mode is also formed between the layers. By means of the mode, the throughput rate during image processing is remarkably improved, and the image processing parallelism degree and speed and the computation performance of the deep learning model are improved. Further provided are an image processing device and system, which have the same beneficial effects as the above image processing method.

First claim

What is claimed is:

1. An image processing method, applied to a deep learning model, wherein a cache queue is provided in front of each layer of the deep learning model, wherein the image processing method comprises:
dividing a received to-be-processed image into a plurality of channel data and storing the plurality of channel data into the cache queue in front of a first layer of the deep learning model; and
obtaining a to-be-processed data corresponding to a plurality of channels from the cache queue in front of at least one layer of the deep learning model, and calling a plurality of computation tasks corresponding to the at least one layer to compute weight parameters and corresponding to-be-processed data in the plurality of channels, and storing computation result into the cache queue behind the at least one layer.

2. The image processing method according to claim 1 , wherein, a number of the computation tasks is equal to a number of channels of the at least one layer.

3. The image processing method according to claim 1 , wherein, a number N of the computation tasks is smaller than a number M of channels of the at least one layer.

4. The image processing method according to claim 3 , wherein, obtaining the to-be-processed data corresponding to the plurality of channels from the cache queue in front of the at least one layer, and calling the plurality of computation tasks corresponding to the at least one layer to compute the weight parameters and the corresponding to-be-processed data in the plurality of channels, comprises:
obtaining the to-be-processed data corresponding to N channels from the cache queue in front of the at least one layer, and calling N computation tasks corresponding to the at least one layer to compute the weight parameters and the corresponding to-be-processed data in the N channels in one-to-one correspondence; and
when a computation task that has completed the computation appears, the to-be-processed data corresponding to uncomputed channels is obtained from the cache queue in front of the at least one layer, and the computation task that has completed the computation is called to continue to compute the weight parameters and the corresponding to-be-processed data in the uncomputed channels until all the channels in the at least one layer have been computed.

5. The image processing method according to claim 1 , wherein, at least one of the computation tasks is a multiplication computation.

6. The image processing method according to claim 1 , wherein, after obtaining the to-be-processed data corresponding to the plurality of channels from the cache queue in front of the at least one layer, the method further comprises:
sending read information of the to-be-processed data corresponding to a previous layer, so that a next to-be-processed image corresponding to the previous layer is processed, and the computation result is saved to a storage position of the to-be-processed data corresponding to the read information of the to-be-processed data.

7. An image processing system, applied to a deep learning model, wherein, a cache queue is provided in front of each layer of the deep learning model, and the image processing system comprises:
a memory for storing computer program; and
a processor for performing steps of the image processing method according to claim 1 when executing the computer program.

8. An image processing system, applied to a deep learning model, wherein a cache queue is provided in front of each layer of the deep learning model, and the image processing system comprises:
a memory for storing computer program; and
a processor for performing steps of the image processing method according to claim 2 when executing the computer program.

9. An image processing system, applied to a deep learning model, wherein a cache queue is provided in front of each layer of the deep learning model, and the image processing system comprises:
a memory for storing computer program; and
a processor for performing steps of the image processing method according to claim 3 when executing the computer program.

10. An image processing system, applied to a deep learning model, wherein a cache queue is provided in front of each layer of the deep learning model, and the image processing system comprises:
a memory for storing computer program; and
a processor for performing steps of the image processing method according to claim 4 when executing the computer program.

11. An image processing system, applied to a deep learning model, wherein a cache queue is provided in front of each layer of the deep learning model, and the image processing system comprises:
a memory for storing computer program; and
a processor for performing steps of the image processing method according to claim 5 when executing the computer program.

12. An image processing system, applied to a deep learning model, wherein a cache queue is provided in front of each layer of the deep learning model, and the image processing system comprises:
a memory for storing computer program; and
a processor for performing steps of the image processing method according to claim 6 when executing the computer program.

13. The image processing method according to claim 2 , wherein, after obtaining the to-be-processed data corresponding to the plurality of channels from the cache queue in front of the at least one layer, the method further comprises:
sending read information of the to-be-processed data corresponding to a previous layer, so that a next to-be-processed image corresponding to the previous layer is processed, and the computation result is saved to a storage position of the to-be-processed data corresponding to the read information of the to-be-processed data.

14. The image processing method according to claim 3 , wherein, after obtaining the to-be-processed data corresponding to the plurality of channels from the cache queue in front of the at least one layer, the method further comprises:
sending read information of the to-be-processed data corresponding to previous layer, so that a next to-be-processed image corresponding to the previous layer is processed, and the computation result is saved to a storage position of the to-be-processed data corresponding to the read information of the to-be-processed data.

15. The image processing method according to claim 4 , wherein, after obtaining the to-be-processed data corresponding to the plurality of channels from the cache queue in front of the at least one layer, the method further comprises:
sending read information of the to-be-processed data corresponding to a previous layer, so that a next to-be-processed image corresponding to the previous layer is processed, and the computation result is saved to a storage position of the to-be-processed data corresponding to the read information of the to-be-processed data.

16. The image processing method according to claim 5 , wherein, after obtaining the to-be-processed data corresponding to the plurality of channels from the cache queue in front of the at least one layer, the method further comprises:
sending read information of the to-be-processed data corresponding to a previous layer, so that a next to-be-processed image corresponding to the previous layer is processed, and the computation result is saved to a storage position of the to-be-processed data corresponding to the read information of the to-be-processed data.

Assignees

• Inspur Electronic Information Industry Co Ltd

Inventors

• Gao Kai
• Guo Zhenhua
• Cao Fang

Classifications

• G06F9/546

  Message passing systems or structures, e.g. queues · CPC title

• G06T1/60

  Memory management · CPC title

• G06N3/063

  using electronic means · CPC title

• G06F2212/454

  Vector or matrix data · CPC title

• G06F9/4881 Primary

  Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues · CPC title