Information processing method and terminal device

1. An information processing method applied to a computation circuit, wherein the computation circuit includes a communication circuit and operation circuit, and the method comprises:

controlling, by the computation circuit, the communication circuit to obtain a target image to be processed, wherein the target image includes a target object to be processed; and

controlling, by the computation circuit, the operation circuit to obtain and call an operation instruction to perform contour detection and extraction on the target object in the target image, so as to obtain a contour output result, wherein

the contour output result is an edge contour of the target object, and the operation instruction is a pre-stored instruction for contour detection and extraction,

wherein the operation circuit includes a primary operation module and a plurality of secondary operation modules, wherein the primary operation module is interconnected with the plurality of secondary operation modules by an interconnection module, and when the operation instruction is a convolution operation instruction,

the calling the operation instruction to perform contour detection and extraction on the target object in the target image includes:

controlling, by the computation circuit, the secondary operation modules to implement a convolution operation of input data and a convolution kernel in a convolutional neural network algorithm, wherein the input data is the target image and the convolutional neural network algorithm corresponds to the convolution operation instruction,

controlling, by the computation circuit, the interconnection module to implement data transfer between the primary operation module and the secondary operation modules, before a forward operation of a neural network fully connected layer starts, transferring, by the primary operation module, the input data to each secondary operation module through the interconnection module, and after computing output scalars by the secondary operation modules, splicing, by the interconnection module, the output scalars of the respective secondary operation modules stage by stage to obtain an intermediate vector, and sending the intermediate vector back to the primary operation module,

controlling, by the computation circuit, the primary operation module to splice intermediate vectors corresponding to all input data into an intermediate result for subsequent operations, and

controlling, by the computation circuit, the primary operation module to add bias data to the intermediate result, and then performing an activation operation.

2. The method of claim 1 , wherein the computation circuit further includes a register circuit and a controller circuit, and the controlling, by the computation circuit, the operation circuit to obtain the operation instruction to perform contour detection and extraction on the target image, so as to obtain the contour output result includes:

controlling, by the computation circuit, the controller circuit to fetch a first operation instruction from the register circuit, and sending, by the computation circuit, the first operation instruction to the operation circuit, and

controlling, by the computation circuit, the operation circuit to call the first operation instruction to perform feature extraction on the target image to obtain a feature image, and

controlling, by the computation circuit, the operation circuit to extract the edge contour of the target object from the feature image as the contour output result.

3. The method of claim 2 , wherein the controlling, by the computation circuit, the operation circuit to extract the edge contour of the target object from the feature image as the contour output result includes:

controlling, by the computation circuit, the operation circuit to identify the target object in the feature image to extract the edge contour of the target object as the contour output result, or

controlling, by the computation circuit, the operation circuit to identify and extract an edge of the target object in the feature image to obtain the contour output result.

4. The method of claim 2 , wherein the controlling, by the computation circuit, the operation circuit to call the first operation instruction to perform feature extraction on the target image to obtain the feature image includes:

controlling, by the computation circuit, the operation circuit to perform feature extraction on the target image based on an operation instruction set of at least one thread to obtain the feature image, wherein the operation instruction set includes at least one first operation instruction, and an order of calling the first operation instruction in the operation instruction set is customized by a user side or a terminal side.

5. The method of claim 1 , wherein the controlling, by the computation circuit, the communication circuit to obtain the target image to be processed includes:

controlling, by the computation circuit, the communication circuit to obtain an original image to be processed input by a user, and

controlling, by the computation circuit, the operation circuit to pre-process the original image to obtain the target image to be processed, wherein the pre-processing includes one or more of the following processing: translation, scaling transformation, non-linear transformation, normalization, format conversion, data deduplication, processing of data exception, and data missing filling.

6. The method of claim 1 , wherein the computation circuit further includes a data access circuit and a storage medium, and

the computation circuit controls the operation circuit to send the contour output result to the data access circuit and store the result in the storage medium.

7. The method of claim 1 , wherein the primary operation module includes a first operation circuit, wherein the first operation circuit includes a vector addition circuit and an activation circuit,

the controlling, by the computation circuit, the primary operation module to add bias data to the intermediate result, and then performing an activation operation include:

controlling, by the computation circuit, the vector addition circuit to implement a bias addition operation of a convolutional neural network operation and perform element-wise addition on bias data and the intermediate result to obtain a bias result, and

controlling, by the computation circuit, the activation circuit to perform an activation function operation on the bias result.

8. The method of claim 1 , wherein each secondary operation module includes a second operation circuit, wherein the secondary operation module includes a vector multiplication circuit and an accumulation circuit,

the controlling, by the computation circuit, the secondary operation modules to perform a convolution operation of input data and a convolution kernel in a convolutional neural network algorithm includes:

controlling, by the computation circuit, the vector multiplication circuit to perform a vector multiplication operation of the convolution operation, and

controlling, by the computation circuit, the accumulation circuit to perform an accumulation operation of the convolution operation.

9. The method of claim 1 , wherein the computation circuit controls the plurality of secondary operation modules to compute respective output scalars in parallel by using the same input data and respective convolution kernels.

10. A computation circuit, comprising a communication circuit and an operation circuit, wherein

the communication circuit is configured to obtain a target image to be processed, wherein the target image includes a target object to be processed; and

the operation circuit is configured to call an operation instruction to perform contour detection and extraction on the target object in the target image, so as to obtain a contour output result, wherein

the contour output result is an edge contour of the target object, and the operation instruction is a pre-stored instruction for contour detection and extraction,

wherein the operation circuit includes a primary operation module and a plurality of secondary operation modules, wherein the primary operation module is interconnected with the plurality of secondary operation modules by an interconnection module, and when the operation instruction is a convolution operation instruction,

the secondary operation modules are configured to implement a convolutional operation of input data and a convolution kernel in a convolution neural network algorithm, wherein the input data is the target image and the convolution neural network algorithm corresponds to the convolutional operation instruction,

the interconnection module is configured to implement data transfer between the primary operation module and the secondary operation modules, before a forward operation of a neural network fully connected layer starts, the primary operation module sends the input data to each secondary operation module through the interconnection module, and after computing output scalars by the secondary operation modules, the interconnection module splices the output scalars of the respective secondary operation modules stage by stage into an intermediate vector and sends the intermediate vector back to the primary operation module, and

the primary operation module is configured to splice intermediate vectors corresponding to all input data into an intermediate result, and perform subsequent operations on the intermediate result, wherein

the primary operation module is configured to add bias data to the intermediate result, and then perform an activation operation.

11. The computation circuit of claim 10 , further comprising a register circuit and a controller circuit, wherein

the controller circuit is configured to fetch a first operation instruction from the register circuit, and send the first operation instruction to the operation circuit, and

the operation circuit is configured to call the first operation instruction to perform feature extraction on the target image to obtain a feature image, and

the operation circuit is configured to extract the edge contour of the target object from the feature image as the contour output result.

12. The computation circuit of claim 11 , wherein

the operation circuit is configured to perform feature extraction on the target image based on an operation instruction set of at least one thread to obtain the feature image, wherein the operation instruction set includes at least one first operation instruction, and an order of calling the first operation instruction in the operation instruction set is customized by a user side or a terminal side, and/or

the operation circuit is configured to identify the target object in the feature image to extract the edge contour of the target object as the contour output result, or the operation circuit is configured to identify and extract the edge of the target object in the target image to obtain the contour output result.

13. The computation circuit of claim 10 , wherein

the communication circuit is configured to obtain an original image to be processed input by a user,

the operation circuit is configured to pre-process the original image to obtain the target image to be processed, wherein the pre-processing includes one or more of the following processing: translation, scaling transformation, non-linear transformation, normalization, format conversion, data deduplication, processing of data exception, and data missing filling.

14. The computation circuit of claim 10 , wherein the plurality of secondary modules use the same input data and respective convolution kernels to compute respective output scalars in parallel.