TWI805005B - Method for detecting images, electronic device and storage medium - Google Patents

Method for detecting images, electronic device and storage medium Download PDF

Info

Publication number
TWI805005B
TWI805005B TW110136617A TW110136617A TWI805005B TW I805005 B TWI805005 B TW I805005B TW 110136617 A TW110136617 A TW 110136617A TW 110136617 A TW110136617 A TW 110136617A TW I805005 B TWI805005 B TW I805005B
Authority
TW
Taiwan
Prior art keywords
node
data
neural network
network model
data pair
Prior art date
Application number
TW110136617A
Other languages
Chinese (zh)
Other versions
TW202316319A (en
Inventor
顏健武
Original Assignee
鴻海精密工業股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 鴻海精密工業股份有限公司 filed Critical 鴻海精密工業股份有限公司
Priority to TW110136617A priority Critical patent/TWI805005B/en
Publication of TW202316319A publication Critical patent/TW202316319A/en
Application granted granted Critical
Publication of TWI805005B publication Critical patent/TWI805005B/en

Links

Images

Landscapes

  • Image Analysis (AREA)
  • Image Processing (AREA)
  • Holo Graphy (AREA)

Abstract

The present application provides a method for detecting images, an electronic device and a storage medium. The method includes: obtaining a neural network model, which includes n operators; mapping the neural network model to a singly linked list structure, which includes n nodes; establishing a data pair vector corresponding to the singly linked list structure based on an output and an input of each node; determining operator subgroups in the neural network model based on the data pair vector; inputting a target image to the neural network model, recording input data and output data of the operator subgroups; detecting the target image based on the input data and the output data, and outputting a detection result. By utilizing the present application, an image detection result can be quickly obtained.

Description

圖像檢測方法、電子設備及存儲介質 Image detection method, electronic device and storage medium

本申請涉及影像處理技術領域,尤其涉及一種圖像檢測方法、電子設備及存儲介質。 The present application relates to the technical field of image processing, and in particular to an image detection method, electronic equipment and a storage medium.

在實際應用中,在藉由安裝在電子設備中的神經網路模型進行圖像檢測時,圖像的特徵資料會在運算元之間進行傳遞。一般情況下,需要記錄每個運算元的輸入資料與輸出資料,並藉由所述每個運算元的輸入資料與輸出資料執行對所述圖像的進一步檢測。然而,有些複雜的神經網路模型中的運算元較多,需要統計的運算元的輸入資料與輸出資料龐大。如此必然會造成電子設備存在記憶體佔用過大,CPU處理慢,檢測圖像速度較慢的情況。 In practical applications, when an image is detected by a neural network model installed in an electronic device, the characteristic data of the image will be transmitted between the operating units. In general, it is necessary to record the input data and output data of each operation element, and perform further detection on the image based on the input data and output data of each operation element. However, some complex neural network models have many operands, and the input data and output data of the operands that need to be counted are huge. This will inevitably cause the electronic device to have excessive memory usage, slow CPU processing, and slow image detection speed.

鑒於以上內容,有必要提供一種圖像檢測方法、電子設備及存儲介質,能快速地輸出圖像檢測結果。 In view of the above, it is necessary to provide an image detection method, an electronic device and a storage medium, which can quickly output image detection results.

本申請提供一種圖像檢測方法,所述方法包括:獲取神經網路模型,其中,所述神經網路模型包括n個運算元;基於所述n個運算元之間的邏輯運算關係,將所述神經網路模型映射為單鏈表結構,其中,所述單鏈表結構包括n個節點;掃描所述單鏈表結構中的每個節點,統計所述每個節點在所述神經 網路模型中的輸出量與輸入量;基於每個節點的輸出量和輸入量建立所述單鏈表結構對應的資料對向量,其中,所述資料對向量為[[a1,b1],[a2,b2]...[ai,bi]...[an,bn]],ai為每個節點的輸出量,bi為每個節點的輸入量;基於所述資料對向量確定所述神經網路模型中的運算元子組;輸入靶心圖表像至所述神經網路模型,記錄所述運算元子組的輸入資料與輸出資料;基於所述輸入資料與輸出資料利用所述神經網路模型對所述靶心圖表像進行檢測,並輸出檢測結果。 The present application provides an image detection method, the method comprising: obtaining a neural network model, wherein the neural network model includes n operands; based on the logical operation relationship between the n operands, the The neural network model is mapped to a single-linked list structure, wherein the single-linked list structure includes n nodes; each node in the single-linked list structure is scanned, and statistics are made on each node in the neural network. The output and input in the network model; based on the output and input of each node, the data pair vector corresponding to the single linked list structure is established, wherein the data pair vector is [[a1, b1], [ a2, b2]...[ai, bi]...[an, bn]], ai is the output of each node, and bi is the input of each node; determine the neural network based on the data pair vector A subgroup of operands in the network model; inputting a bullseye image to the neural network model, recording input data and output data of the subgroup of operands; utilizing the neural network based on the input data and output data The model detects the bull's-eye image and outputs the detection result.

在一種可能的實現方式中,所述單鏈表中的n節點分別為OP1,OP2,...OPi,...,OPn,節點OPi的輸出量ai表示所述節點OPi對應的運算元在所述神經網路模型中將資料向外進行傳遞時,接收所述資料的運算元的數量;所述節點OPi的輸入量bi表示所述節點OPi對應的運算元在所述神經網路模型中接收其他運算元傳遞的資料時,傳遞所述資料的所述其他運算元的數量。 In a possible implementation, the n nodes in the singly linked list are OP1, OP2, ... OPi, ..., OPn, and the output ai of the node OPi indicates that the operand corresponding to the node OPi is in When the data is transmitted in the neural network model, the number of operands receiving the data; the input quantity bi of the node OPi indicates that the operand corresponding to the node OPi is in the neural network model When receiving the data transferred by other operands, the quantity of the other operands that transfer the data.

在一種可能的實現方式中,基於所述資料對向量確定所述神經網路模型中的運算元子組包括:根據所述資料對向量確定多個節點子集;根據所述多個節點子集確定所述神經網路模型中的運算元子組。 In a possible implementation manner, determining the operator subset in the neural network model based on the data pair vector includes: determining a plurality of node subsets according to the data pair vector; A subset of operands in the neural network model is determined.

在一種可能的實現方式中,根據所述資料對向量確定多個節點子集包括:S51、從[a1,b1]開始遍歷所述資料對向量中的資料對;S52、判斷所述資料對是否滿足壓入堆疊條件,若所述資料對不滿足壓入堆疊條件,執行S54;S53、若所述資料對滿足壓入堆疊條件,將所述資料對壓入堆疊中;S54、判斷所述資料對是否滿足彈出堆疊條件,若所述資料對不滿足彈出堆疊條件,執行S57;S55、若所述資料對滿足彈出堆疊條件,彈出當前堆疊中最頂層的資料對,並統計所述堆疊中剩餘資料對的數量m;S56、確定所述堆疊中最頂層的資料對對應的起始節點,以及確定滿足所述彈出條件的資料對對應的末尾節點,設 定從所述起始節點至所述末尾節點的所有節點構成的子集為m+1級節點子集;S57、繼續遍歷所述資料對向量中的資料對,重複執行S52至S57,直到所述資料對向量中的資料對被全部遍歷;S58、結束遍歷。 In a possible implementation manner, determining multiple node subsets according to the data pair vector includes: S51, traversing the data pairs in the data pair vector starting from [a1, b1]; S52, judging whether the data pair Satisfy the push-in stacking condition, if the data pair does not meet the push-in stacking condition, execute S54; S53, if the data pair meets the push-in stacking condition, push the data pair into the stack; S54, judge the data Whether the pop-up stacking condition is satisfied, if the data pair does not meet the pop-up stacking condition, execute S57; S55, if the data pair meets the pop-up stacking condition, pop up the topmost data pair in the current stack, and count the remaining The number of data pairs m; S56, determine the start node corresponding to the topmost data pair in the stack, and determine the end node corresponding to the data pair that meets the pop-up condition, set Determine the subset formed by all nodes from the start node to the end node as the m+1 level node subset; S57, continue to traverse the data pairs in the data pair vector, and repeatedly execute S52 to S57 until all The data pairs in the data pair vector are all traversed; S58, end the traverse.

在一種可能的實現方式中,所述根據所述多個節點子集確定所述神經網路模型中的運算元子組包括:遍歷所述多個節點子集;若所述多個節點子集中的任意一個節點子集中的節點與其他節點子集中的節點都不重複,確定每個節點子集為所述神經網路模型中的一個運算元子組。 In a possible implementation manner, the determining the operator subset in the neural network model according to the multiple node subsets includes: traversing the multiple node subsets; if the multiple node subsets The nodes in any one of the node subsets are not repeated with the nodes in other node subsets, and each node subset is determined to be a subset of operands in the neural network model.

在一種可能的實現方式中,根據所述多個節點子集確定所述神經網路模型中的運算元子組還包括:遍歷所述多個節點子集,得到多個第一節點子集和多個第二節點子集,其中,所述第一節點子集中的全部節點真包含於所述第二節點子集中;去除所述第二節點子集中的所述第一節點子集的全部節點;設定所述第二節點子集中剩餘的節點集合為所述神經網路模型的運算元子組。 In a possible implementation manner, determining the operator subset in the neural network model according to the multiple node subsets further includes: traversing the multiple node subsets to obtain multiple first node subsets and A plurality of second subsets of nodes, wherein all nodes in the first subset of nodes are actually included in the second subset of nodes; removing all nodes of the first subset of nodes in the second subset of nodes ; Set the remaining node set in the second node subset as the operator subset of the neural network model.

在一種可能的實現方式中,根據所述多個節點子集確定所述神經網路模型中的運算元子組還包括:設定所述多個第一節點子集為所述神經網路模型的運算元子組。 In a possible implementation manner, determining the operator subset in the neural network model according to the multiple node subsets further includes: setting the multiple first node subsets as the neural network model's Subgroup of operands.

在一種可能的實現方式中,若資料對[ai,bi]中的bi 2,確定所述資料對[ai,bi]滿足彈出堆疊條件;若資料對[ai,bi]中的ai 2,確定所述資料對[ai,bi]滿足壓入堆疊條件。 In a possible implementation, if bi 2 in the data pair [ai, bi], determine that the data pair [ai, bi] satisfies the pop-up stacking condition; if ai 2 in the data pair [ai, bi], determine The data pair [ai, bi] satisfies the push-in stacking condition.

本申請還提供一種電子設備,所述電子設備包括處理器和記憶體,所述處理器用於執行所述記憶體中存儲的電腦程式時實現所述的圖像檢測方法。 The present application also provides an electronic device, the electronic device includes a processor and a memory, and the processor is configured to implement the image detection method when executing a computer program stored in the memory.

本申請還提供一種電腦可讀存儲介質,所述電腦可讀存儲介質上存 儲有電腦程式,所述電腦程式被處理器執行時實現所述的圖像檢測方法。 The present application also provides a computer-readable storage medium, the computer-readable storage medium stores A computer program is stored, and when the computer program is executed by the processor, the image detection method is realized.

本申請公開的圖像檢測方法、電子設備及存儲介質,能快速地輸出圖像檢測結果。 The image detection method, electronic equipment and storage medium disclosed in the present application can quickly output image detection results.

S1~S7:步驟 S1~S7: steps

S51~S58:步驟 S51~S58: steps

5:電子設備 5: Electronic equipment

51:記憶體 51: memory

52:處理器 52: Processor

53:電腦程式 53: Computer program

54:通訊匯流排 54: communication bus

OP1~OP15:節點 OP1~OP15: Node

圖1是本申請公開的一種圖像檢測方法的較佳實施例的流程圖。 Fig. 1 is a flowchart of a preferred embodiment of an image detection method disclosed in this application.

圖2是本申請公開的一種圖像檢測方法的步驟S5的詳細步驟流程圖。 FIG. 2 is a detailed flow chart of step S5 of an image detection method disclosed in the present application.

圖3是本申請公開的一種示例性單鏈表結構。 FIG. 3 is an exemplary singly linked list structure disclosed in this application.

圖4是本申請公開的一種示例性節點資料傳輸圖。 Fig. 4 is an exemplary node data transmission diagram disclosed in the present application.

圖5是本申請實現圖像檢測方法的較佳實施例的電子設備的結構示意圖。 Fig. 5 is a schematic structural diagram of an electronic device implementing a preferred embodiment of the image detection method of the present application.

為了使本申請的目的、技術方案和優點更加清楚,下面結合附圖和具體實施例對本申請進行詳細描述。 In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be described in detail below in conjunction with the accompanying drawings and specific embodiments.

如圖1所示,是本申請圖像檢測方法的較佳實施例的流程圖。根據不同的需求,該流程圖中步驟的順序可以改變,某些步驟可以省略。 As shown in FIG. 1 , it is a flow chart of a preferred embodiment of the image detection method of the present application. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted.

所述圖像檢測方法應用於一個或者多個電子設備5中,所述電子設備5是一種能夠按照事先設定或存儲的指令,自動進行數值計算和/或資訊處理的設備,其硬體包括但不限於微處理器、專用積體電路(Application Specific Integrated Circuit,ASIC)、可程式設計閘陣列(Field-Programmable Gate Array,FPGA)、數位訊號處理器(Digital Signal Processor,DSP)、嵌入式設備等。 The image detection method is applied to one or more electronic devices 5, and the electronic device 5 is a device capable of automatically performing numerical calculation and/or information processing according to preset or stored instructions, and its hardware includes but Not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital signal processors (Digital Signal Processor, DSP), embedded devices, etc. .

所述電子設備5可以是任何一種可與用戶進行人機交互的電子產品,例如,個人電腦、平板電腦、智慧手機、個人數位助理(Personal Digital Assistant, PDA)、遊戲機、互動式網路電視(Internet Protocol Television,IPTV)、智慧式穿戴式設備等。 The electronic device 5 can be any electronic product capable of man-machine interaction with the user, for example, a personal computer, a tablet computer, a smart phone, a personal digital assistant (Personal Digital Assistant, PDA), game consoles, interactive Internet TV (Internet Protocol Television, IPTV), smart wearable devices, etc.

所述電子設備5還可以包括網路設備和/或使用者設備。其中,所述網路設備包括,但不限於單個網路服務器、多個網路服務器組成的伺服器組或基於雲計算(Cloud Computing)的由大量主機或網路服務器構成的雲。 The electronic equipment 5 may also include network equipment and/or user equipment. Wherein, the network device includes, but is not limited to, a single network server, a server group composed of multiple network servers, or a cloud composed of a large number of hosts or network servers based on Cloud Computing.

所述電子設備5所處的網路包括但不限於互聯網路、廣域網路、都會區網路、局域網路、虛擬私人網路(Virtual Private Network,VPN)等。 The network where the electronic device 5 is located includes, but is not limited to, the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (Virtual Private Network, VPN) and the like.

步驟S1、獲取神經網路模型,其中,所述神經網路模型包括n個運算元,其中,n表示正整數。 Step S1. Obtain a neural network model, wherein the neural network model includes n operands, where n represents a positive integer.

在本實施方式中,所述神經網路模型可以為直接從網路端獲取到的神經網路模型,也可以為從所述網路端獲取到的神經網路模型進行優化以後的神經網路模型。對神經網路模型進行優化可以理解為對所述神經網路模型進行運算元融合,網路剪枝,模型量化,網路切割等操作。在本實施方式中,所述神經網路模型包括n個運算元,分別為op1,op2,...,opn。 In this embodiment, the neural network model may be a neural network model obtained directly from the network end, or may be a neural network model after optimization of the neural network model obtained from the network end. Model. Optimizing the neural network model can be understood as performing operations such as operator fusion, network pruning, model quantization, and network cutting on the neural network model. In this embodiment, the neural network model includes n operands, which are respectively op1, op2, . . . , opn.

步驟S2、基於所述n個運算元之間的邏輯運算關係,將所述神經網路模型映射為單鏈表結構,其中,所述單鏈表結構包括n個節點。 Step S2: Map the neural network model into a singly linked list structure based on the logical operation relationship between the n operands, wherein the singly linked list structure includes n nodes.

為了後續統計每個運算元的輸入與輸出資料,需要將所述神經網路模型映射為單鏈表結構。 In order to subsequently count the input and output data of each operator, it is necessary to map the neural network model into a singly linked list structure.

在本實施方式中,所述單鏈表結構包括多個節點,並且所述多個節點單向排列。排列在所述單鏈表結構的起始位置的節點為所述單鏈表結構的頭部,排列在所述單鏈表結構的截止位置的節點為所述單鏈表結構的尾部。在訪問所述單鏈表結構時,需要根據單鏈表的單向順序從頭部開始讀取,到尾部結束。 In this implementation manner, the singly linked list structure includes a plurality of nodes, and the plurality of nodes are arranged in a unidirectional manner. The node arranged at the start position of the singly linked list structure is the head of the singly linked list structure, and the node arranged at the end position of the singly linked list structure is the tail of the singly linked list structure. When accessing the singly linked list structure, it is necessary to read from the head to the tail according to the unidirectional order of the singly linked list.

在本實施方式中,所述n個運算元之間的邏輯運算關係可以理解為運算元之間的資料傳遞關係。例如,所述神經網路模型包括運算元A,運算元B,運算元C以及運算元D。其中,運算元A的輸出資料為運算元B與運算元C的輸入資料,運算元B的輸出資料與運算元C的輸出資料為運算元D的輸入資料。如此,可以得到的單鏈表結構包括A─>B─>C─>D,或者A─>C─>B─>D。 In this embodiment, the logical operation relationship between the n operands can be understood as a data transfer relationship between the operands. For example, the neural network model includes an operator A, an operator B, an operator C, and an operator D. Wherein, the output data of the operator A is the input data of the operator B and the operator C, and the output data of the operator B and the output data of the operator C are the input data of the operator D. In this way, the single linked list structure that can be obtained includes A─>B─>C─>D, or A─>C─>B─>D.

在本實施方式中,所述單鏈表結構包括n個節點,所述n個節點分別為OP1,OP2,...,OPn,所述n個節點一一對應於所述神經網路模型中的每一個運算元。基於所述n個運算元之間的依賴關係,可以得到多個單鏈表結構。在本實施方式中,選取其中一種單鏈表結構OP1─>OP2─>...─>OPn進行後續的說明。例如,圖3所示的單鏈表結構。 In this embodiment, the singly linked list structure includes n nodes, and the n nodes are respectively OP1, OP2, ..., OPn, and the n nodes correspond one by one to the each operand of . Based on the dependency relationship between the n operands, multiple singly linked list structures can be obtained. In this embodiment, one of the singly linked list structures OP1─>OP2─>...─>OPn is selected for subsequent description. For example, the singly linked list structure shown in Figure 3.

步驟S3、掃描所述單鏈表結構中的每個節點,統計所述每個節點在所述神經網路模型中的輸出量與輸入量。 Step S3, scanning each node in the singly linked list structure, and counting the output and input of each node in the neural network model.

為了後續獲得運算元子組,需要統計所述每個節點在所述神經網路模型中的輸出量與輸入量。 In order to subsequently obtain the subgroup of operands, it is necessary to count the output and input of each node in the neural network model.

在本實施方式中,所述輸出量表示節點對應的運算元在所述神經網路模型中將資料向外進行傳遞時,接收所述資料的運算元的數量;所述輸入量表示節點對應的運算元在所述神經網路模型中接收其他運算元傳遞的資料時,傳遞所述資料的運算元的數量。 In this embodiment, the output amount represents the number of operation elements that receive the data when the operation element corresponding to the node transmits the data to the outside in the neural network model; the input amount represents the number of operation elements corresponding to the node. When the operation unit receives the data delivered by other operation units in the neural network model, the number of the operation unit that transmits the data.

示例性的,圖4中的節點OP4的輸入量為1,輸出量為4。 Exemplarily, the input quantity of the node OP4 in FIG. 4 is 1, and the output quantity is 4.

步驟S4、基於每個節點的輸出量和輸入量建立所述單鏈表結構對應的資料對向量,其中,所述資料對向量為[[a1,b1],[a2,b2]...[ai,bi]...[an,bn]],ai為每個節點的輸出量,bi為每個節點的輸入量,其中,ai表示正整數, bi表示正整數。 Step S4, based on the output and input of each node, establish a data pair vector corresponding to the singly linked list structure, wherein the data pair vector is [[a1, b1], [a2, b2]...[ ai, bi]...[an, bn]], ai is the output of each node, bi is the input of each node, where ai represents a positive integer, bi represents a positive integer.

在本實施方式中,所述資料對向量為根據節點OP1,OP2,...,以及OPn的輸出量與輸入量建立的二維陣列,其中[ai,bi]表示節點OPi的輸出量與輸入量資料。 In this embodiment, the data pair vector is a two-dimensional array established according to the output and input of the nodes OP1, OP2, ..., and OPn, where [ai, bi] represents the output and input of the node OPi Quantitative data.

步驟S5、基於所述資料對向量確定所述神經網路模型中的運算元子組。 Step S5, determining a subset of operands in the neural network model based on the data pair vector.

在本實施方式中,所述電子設備在藉由神經網路模型進行圖像檢測時,需要記錄每個運算元的輸入資料與輸出資料。另外,所述電子設備需要記錄的資料還包括一些無效資料。如此,所述電子設備需要記錄的資料量龐大,可能導致所述電子設備的記憶體佔用過大,CPU處理慢。為了提高電子設備的處理能力,快速得到圖像檢測結果,可以藉由本申請提供的方法查找神經網路模型中的運算元子組。如此,可以只記錄運算元子組的輸入資料與輸出資料,就可以進行圖像檢測的下一步處理。 In this embodiment, when the electronic device performs image detection through the neural network model, it needs to record the input data and output data of each operation unit. In addition, the data to be recorded by the electronic device also includes some invalid data. In this way, the electronic device needs to record a huge amount of data, which may lead to excessive memory usage and slow CPU processing of the electronic device. In order to improve the processing capability of electronic equipment and quickly obtain image detection results, the method provided by this application can be used to search for the subgroup of operands in the neural network model. In this way, only the input data and output data of the operator subgroup can be recorded, and the next step of image detection can be performed.

具體地,基於所述資料對向量確定所述神經網路模型中的運算元子組包括:根據所述資料對向量確定多個節點子集;根據所述多個節點子集確定所述神經網路模型中的運算元子組。 Specifically, determining the operator subset in the neural network model based on the data pair vector includes: determining a plurality of node subsets according to the data pair vector; determining the neural network model according to the plurality of node subsets. A subgroup of operands in the path model.

參閱圖2所示,是本申請公開的一種圖像檢測方法的步驟S5的詳細步驟流程圖。根據不同的需求,該流程圖中步驟的順序可以改變,某些步驟可以省略。 Referring to FIG. 2 , it is a detailed flow chart of step S5 of an image detection method disclosed in the present application. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted.

S51、從[a1,b1]開始遍歷所述資料對向量中的資料對;S52、判斷所述資料對是否滿足壓入堆疊條件,其中所述滿足壓入堆疊條件為資料對[ai,bi]中的ai>=2,若所述資料對不滿足壓入堆疊條件,執行S54; S53、若所述資料對滿足壓入堆疊條件,將所述資料對壓入堆疊中;S54、判斷所述資料對是否滿足彈出堆疊條件,其中所述滿足彈出堆疊條件為資料對[ai,bi]中的bi>=2,若所述資料對不滿足彈出堆疊條件,執行S57;S55、若所述資料對滿足彈出堆疊條件,彈出當前堆疊中最頂層的資料對,並統計所述堆疊中剩餘資料對的數量m,其中,m表示正整數;S56、確定所述堆疊中最頂層的資料對對應的起始節點,以及確定滿足所述彈出堆疊條件的資料對對應的末尾節點,設定從所述起始節點至所述末尾節點的所有節點構成的子集為m+1級節點子集;S57、繼續遍歷所述資料對向量中的資料對,重複執行S52至S57,直到所述資料對向量中的資料對被全部遍歷;S58、結束遍歷。 S51. Start traversing the data pairs in the data pair vector from [a1, b1]; S52. Determine whether the data pairs meet the push-in stacking conditions, wherein the data pairs [ai, bi] satisfy the push-in stacking conditions ai>=2, if the data pair does not meet the push stacking condition, execute S54; S53. If the data pair satisfies the push-in stacking condition, push the data pair into the stack; S54. Determine whether the data pair satisfies the pop-up stacking condition, wherein the data pair [ai, bi ] in bi>=2, if the data pair does not meet the pop-up stacking condition, execute S57; S55, if the data pair meets the pop-up stacking condition, pop up the topmost data pair in the current stack, and count the data in the stack The number m of the remaining data pairs, wherein m represents a positive integer; S56, determine the start node corresponding to the topmost data pair in the stack, and determine the end node corresponding to the data pair that satisfies the pop-up stacking condition, set from The subset formed by all nodes from the start node to the end node is a subset of m+1 level nodes; S57, continue to traverse the data pairs in the data pair vector, and repeatedly execute S52 to S57 until the data The data pairs in the pair vector are all traversed; S58, end the traverse.

經過上述步驟S51至步驟S58處理後,得到的多個節點子集會存在兩種情況。一種情況是所述多個節點子集中任意兩個節點子集不存在重複的節點;另一種情況是所述多個節點子集中的某一個子集真包含於其他子集。對於不同的情況,確定所述神經網路模型中的運算元子組的方法也不相同。 After the above steps S51 to S58 are processed, there are two situations in the multiple node subsets obtained. One case is that there are no duplicate nodes in any two node subsets among the plurality of node subsets; the other case is that a certain subset among the plurality of node subsets is really included in other subsets. For different situations, the method for determining the subgroup of operands in the neural network model is also different.

在一實施方式,所述根據所述多個節點子集確定所述神經網路模型中的運算元子組包括:遍歷所述多個節點子集;若所述多個節點子集中的任意一個節點子集中的節點與其他節點子集中的節點都不重複,確定每個節點子集為所述神經網路模型中的一個運算元子組。 In an embodiment, the determining the operator subset in the neural network model according to the plurality of node subsets includes: traversing the plurality of node subsets; if any one of the plurality of node subsets Nodes in the node subset are not repeated with nodes in other node subsets, and each node subset is determined as a subset of operands in the neural network model.

在另一實施方式,所述根據所述多個節點子集確定所述神經網路模型中的運算元子組包括:遍歷所述多個節點子集,得到多個第一節點子集和多個第二節點子集,其中,所述第一節點子集真包含於所述第二節點子集中;去 除所述第二節點子集中的所述第一節點子集的全部節點,將所述第二節點子集中剩餘的節點集合作為所述神經網路模型的運算元子組。 In another embodiment, the determining the operator subsets in the neural network model according to the multiple node subsets includes: traversing the multiple node subsets to obtain multiple first node subsets and multiple a second subset of nodes, wherein the first subset of nodes is actually included in the second subset of nodes; Except all the nodes in the first node subset in the second node subset, the remaining node sets in the second node subset are used as a subset of operands of the neural network model.

需要說明的是,所述第一節點子集與所述第二節點子集中剩餘的節點集合不存在重複的節點。因此需要設定所述多個第一節點子集為所述神經網路模型的運算元子組。 It should be noted that there are no duplicate nodes between the first node subset and the remaining node sets in the second node subset. Therefore, it is necessary to set the plurality of first node subsets as the operator subsets of the neural network model.

示例性的,若單鏈表結構為OP1─>OP2─>...─>OP15,其中所述節點OP1至OP11分別對應於神經網路模型的運算元op1至op15,所述單鏈表中的節點對應的資料對向量為[[2,1],[1,1],[1,2],[4,1],[1,1],[1,1],[1,1],[1,1],[1,1],[1,1],[1,1],[2,1],[1,1],[1,2],[1,4]];從[2,1]開始遍歷所述資料對向量;由於資料對[2,1]滿足壓入堆疊條件,將資料對[2,1]壓入堆疊;由於資料對[1,2]滿足彈出堆疊條件,將當前堆疊中最頂層的資料對[2,1]彈出;統計所述堆疊中的資料對的數量為0,確定節點OP1,OP2,OP3為1級節點子集;繼續遍歷所述資料對向量;由於資料對[4,1]滿足壓入堆疊條件,將資料對[4,1]壓入堆疊;由於資料對[2,1]滿足壓入堆疊條件,將資料對[2,1]壓入堆疊;由於資料對[1,2]滿足彈出堆疊條件,將當前堆疊中最頂層的資料對[2,1]彈出;統計所述堆疊中的資料對的數量為1,確定節點OP12,OP13,OP14為2級節點子集; 繼續遍歷所述資料對向量;由於資料對[1,4]滿足彈出堆疊條件,將當前堆疊中最頂層的資料對[4,1]彈出;統計所述堆疊中的資料對的數量為0,確定節點OP4,OP5,...,OP15為1級節點子集;結束遍歷,確定所述神經網路模型中的運算元子組包括兩個1級運算元子組和一個2級運算元子組。其中,一個所述1級運算元子組包括運算元op1,op2,和op3;另一個所述1級運算元子組包括運算元op4,op5,op6,op7,op8,op9,op10,op11,和op15。所述2級運算元子組包括op12,op13,和op14。 Exemplarily, if the structure of the singly linked list is OP1─>OP2─>...─>OP15, wherein the nodes OP1 to OP11 respectively correspond to the operational elements op1 to op15 of the neural network model, in the singly linked list The data pair vector corresponding to the node is [[2, 1], [1, 1], [1, 2], [4, 1], [1, 1], [1, 1], [1, 1] , [1,1], [1,1], [1,1], [1,1], [2,1], [1,1], [1,2], [1,4]]; Start traversing the data pair vector from [2, 1]; since the data pair [2, 1] satisfies the push-in stacking condition, push the data pair [2, 1] into the stack; because the data pair [1, 2] satisfies the pop-up Stacking conditions, pop up the topmost data pair [2, 1] in the current stack; count the number of data pairs in the stack as 0, and determine that nodes OP1, OP2, and OP3 are the first-level node subsets; continue to traverse the Data pair vector; since the data pair [4, 1] satisfies the push-in stacking condition, push the data pair [4, 1] into the stack; because the data pair [2, 1] satisfies the push-in stacking condition, push the data pair [2, 1] into the stack 1] Push into the stack; since the data pair [1, 2] satisfies the pop-up stacking condition, pop the topmost data pair [2, 1] in the current stack; count the number of data pairs in the stack as 1, and determine the node OP12, OP13, and OP14 are a subset of level 2 nodes; Continue traversing the data pair vector; since the data pair [1, 4] satisfies the pop-up stacking condition, pop the topmost data pair [4, 1] in the current stack; count the number of data pairs in the stack as 0, Determine that the nodes OP4, OP5, ..., OP15 are a first-level node subset; end the traversal, and determine that the operator subgroups in the neural network model include two first-level operator subgroups and one second-level operator subgroup Group. Wherein, one of the 1-level operator subgroups includes the operator op1, op2, and op3; the other 1-level operator subgroup includes the operator op4, op5, op6, op7, op8, op9, op10, op11, and op15. The 2-level operand subset includes op12, op13, and op14.

步驟S6、輸入靶心圖表像至所述神經網路模型,記錄所述運算元子組的輸入資料與輸出資料。 Step S6 , inputting the bull's-eye diagram image to the neural network model, and recording the input data and output data of the operator subgroup.

在神經網路模型進行圖像檢測時,圖像的特徵資料會在運算元之間進行傳遞,需要記錄運算元子組的輸入資料與輸出資料,就可以進行下一步圖像的處理與計算。 When the neural network model performs image detection, the feature data of the image will be transmitted between the operators, and the input data and output data of the operator subgroup need to be recorded, and then the next step of image processing and calculation can be performed.

在本實施方式中,所述運算元子組的輸入資料為所述起始節點對應的起始運算元的輸出資料,所述運算元子組的輸出資料為所述末尾節點對應的末尾運算元的輸入資料。藉由本申請提供的方法可以僅記錄所述神經網路模型中的運算元子組的輸入資料與輸出資料,不需要記錄所述神經網路模型中的所有運算元的輸入資料與輸出資料,從而可以大幅度地減少記錄的資料量。 In this embodiment, the input data of the operand subgroup is the output data of the initial operand corresponding to the start node, and the output data of the operand subgroup is the last operand corresponding to the end node input data. The method provided by the present application can only record the input data and output data of the sub-group of operation elements in the neural network model, and does not need to record the input data and output data of all the operation elements in the neural network model, thus The amount of recorded data can be greatly reduced.

步驟S7、基於所述輸入資料與輸出資料利用所述神經網路模型對所述靶心圖表像進行檢測,並輸出檢測結果。 Step S7 , using the neural network model to detect the bull's-eye image based on the input data and output data, and output a detection result.

在本實施方式中,若需要藉由所述神經網路模型對所述靶心圖表像 中的人臉進行檢測時,需要提取所述靶心圖表像中的特徵。所述特徵主要包括顏色、紋理以及邊緣。因此,所述輸入資料與輸出資料主要包括顏色資料、紋理資料以及邊緣資料。藉由所述輸入資料與輸出資料在所述神經網路模型中運算元子組之間的傳遞完成對靶心圖表像的檢測,並輸出人臉檢測結果。 In this embodiment, if it is necessary to use the neural network model to When detecting the face in the image, it is necessary to extract the features in the bull's-eye image. The features mainly include color, texture and edge. Therefore, the input data and output data mainly include color data, texture data and edge data. Through the transmission of the input data and output data among the operator subgroups in the neural network model, the detection of the bull's-eye image is completed, and the face detection result is output.

需要說明的是,所述電子設備5集成的模組/單元如果以軟體功能單元的形式實現並作為獨立的產品銷售或使用時,可以存儲在一個電腦可讀取存儲介質中。基於這樣的理解,本申請實現上述實施例方法中的全部或部分流程,也可以藉由電腦程式來指令相關的硬體來完成,所述的電腦程式可存儲於一電腦可讀存儲介質中,該電腦程式在被處理器執行時,可實現上述各個方法實施例的步驟。其中,所述電腦程式代碼可以為原始程式碼形式、物件代碼形式、可執行檔或某些中間形式等。所述電腦可讀介質可以包括:能夠攜帶所述電腦程式代碼的任何實體或裝置、記錄介質、隨身碟、移動硬碟、磁碟、光碟、電腦記憶體、唯讀記憶體(Read-Only Memory,ROM)。 It should be noted that, if the integrated modules/units of the electronic device 5 are realized in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, all or part of the processes in the methods of the above embodiments of the present application can also be completed by instructing related hardware through computer programs, and the computer programs can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can realize the steps of the above-mentioned various method embodiments. Wherein, the computer program code may be in the form of original code, object code, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, flash drive, removable hard disk, magnetic disk, optical disk, computer memory, read-only memory (Read-Only Memory) , ROM).

在本申請所提供的幾個實施例中,應該理解到,所揭露的系統,裝置和方法,可以藉由其它的方式實現。例如,以上所描述的裝置實施例僅僅是示意性的,例如,所述模組的劃分,僅僅為一種邏輯功能劃分,實際實現時可以有另外的劃分方式。 In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

所述作為分離部件說明的模組可以是或者也可以不是物理上分開的,作為模組顯示的部件可以是或者也可以不是物理單元,即可以位於一個地方,或者也可以分佈到多個網路單元上。可以根據實際的需要選擇其中的部分或者全部模組來實現本實施例方案的目的。 The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or may also be distributed to multiple networks on the unit. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

另外,在本申請各個實施例中的各功能模組可以集成在一個處理單元中,也可以是各個單元單獨物理存在,也可以兩個或兩個以上單元集成在一 個單元中。上述集成的單元既可以採用硬體的形式實現,也可以採用硬體加軟體功能模組的形式實現。 In addition, each functional module in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one processing unit. in units. The above-mentioned integrated units can be implemented not only in the form of hardware, but also in the form of hardware plus software function modules.

對於本領域技術人員而言,顯然本申請不限於上述示範性實施例的細節,而且在不背離本申請的精神或基本特徵的情況下,能夠以其他的具體形式實現本申請。因此,無論從哪一點來看,均應將實施例看作是示範性的,而且是非限制性的,本申請的範圍由所附請求項而不是上述說明限定,因此旨在將落在請求項的等同要件的含義和範圍內的所有變化涵括在本申請內。不應將請求項中的任何附關聯圖標記視為限制所涉及的請求項。此外,顯然“包括”一詞不排除其他單元或步驟,單數不排除複數。系統請求項中陳述的多個單元或裝置也可以由一個單元或裝置藉由軟體或者硬體來實現。第二等詞語用來表示名稱,而並不表示任何特定的順序。 It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above, but that the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, no matter from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the application is defined by the appended claims rather than the above description, so it is intended to All changes within the meaning and range of equivalents of the elements are embraced in this application. Any attached reference mark in a claim shall not be deemed to limit the claim to which it relates. In addition, it is obvious that the word "comprising" does not exclude other elements or steps, and the singular does not exclude the plural. A plurality of units or devices stated in the system requirements can also be implemented by one unit or device by software or hardware. Secondary terms are used to denote names without implying any particular order.

最後應說明的是,以上實施例僅用以說明本申請的技術方案而非限制,儘管參照較佳實施例對本申請進行了詳細說明,本領域的普通技術人員應當理解,可以對本申請的技術方案進行修改或等同替換,而不脫離本申請技術方案的精神和範圍。 Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application without limitation. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be Make modifications or equivalent replacements without departing from the spirit and scope of the technical solutions of the present application.

S1~S7:步驟 S1~S7: steps

Claims (9)

一種圖像檢測方法,應用在電子設備中,其中,所述圖像檢測方法包括:獲取神經網路模型,其中,所述神經網路模型包括n個運算元,其中,n表示正整數;基於所述n個運算元之間的邏輯運算關係,將所述神經網路模型映射為單鏈表結構,其中,所述單鏈表結構包括n個節點,所述單鏈表中的n個節點分別為OP1,OP2,...OPi,...,OPn,節點OPi的輸出量ai表示所述節點OPi對應的運算元在所述神經網路模型中將資料向外進行傳遞時,所統計到的接收所述資料的運算元的數量;所述節點OPi的輸入量bi表示所述節點OPi對應的運算元在所述神經網路模型中接收其他運算元傳遞的資料時,所統計到的所述其他運算元的數量,其中,ai表示正整數,bi表示正整數;掃描所述單鏈表結構中的每個節點,統計所述每個節點在所述神經網路模型中的輸出量與輸入量;基於每個節點的輸出量和輸入量建立所述單鏈表結構對應的資料對向量,其中,所述資料對向量為[[a1,b1],[a2,b2]...[ai,bi]...[an,bn]];基於所述資料對向量確定所述神經網路模型中的運算元子組;輸入靶心圖表像至所述神經網路模型,記錄所述運算元子組的輸入資料與輸出資料;基於所述輸入資料與輸出資料,利用所述神經網路模型對所述靶心圖表像進行檢測,並輸出檢測結果。 An image detection method used in electronic equipment, wherein the image detection method includes: acquiring a neural network model, wherein the neural network model includes n operands, wherein n represents a positive integer; based on The logical operation relationship between the n operands maps the neural network model into a singly linked list structure, wherein the singly linked list structure includes n nodes, and the n nodes in the singly linked list are respectively OP1, OP2, ...OPi, ..., OPn, and the output ai of the node OPi indicates that when the operator corresponding to the node OPi transmits the data in the neural network model, the statistics The number of operands that receive the data; the input quantity bi of the node OPi represents the number of statistics obtained when the operand corresponding to the node OPi receives the data delivered by other operands in the neural network model The number of other operands, wherein ai represents a positive integer, and bi represents a positive integer; scan each node in the singly linked list structure, and count the output of each node in the neural network model and the input amount; establishing a data pair vector corresponding to the single linked list structure based on the output amount and input amount of each node, wherein the data pair vector is [[a1, b1], [a2, b2]... [ai, bi]...[an, bn]]; determine the operator subgroup in the neural network model based on the data pair vector; input the bullseye image to the neural network model, record the The input data and the output data of the operation element subgroup; based on the input data and the output data, the neural network model is used to detect the bull's-eye diagram image, and output the detection result. 如請求項1所述的圖像檢測方法,其中,基於所述資料對向量確定所述神經網路模型中的運算元子組包括:根據所述資料對向量確定多個節點子集; 根據所述多個節點子集確定所述神經網路模型中的運算元子組。 The image detection method according to claim 1, wherein determining the operator subset in the neural network model based on the data pair vector includes: determining a plurality of node subsets according to the data pair vector; A subset of operands in the neural network model is determined according to the plurality of node subsets. 如請求項2所述的圖像檢測方法,其中,根據所述資料對向量確定多個節點子集包括:步驟1:從[a1,b1]開始遍歷所述資料對向量中的資料對;步驟2:判斷所述資料對是否滿足壓入堆疊條件,若所述資料對不滿足壓入堆疊條件,執行步驟4;步驟3:若所述資料對滿足壓入堆疊條件,將所述資料對壓入堆疊中;所述步驟4:判斷所述資料對是否滿足彈出堆疊條件,若所述資料對不滿足彈出堆疊條件,執行步驟7;步驟5:若所述資料對滿足彈出堆疊條件,彈出當前堆疊中最頂層的資料對,並統計所述堆疊中剩餘資料對的數量m,其中,m表示正整數;步驟6:確定所述堆疊中最頂層的資料對對應的起始節點,以及確定滿足所述彈出堆疊條件的資料對對應的末尾節點,設定從所述起始節點至所述末尾節點的所有節點構成的子集為m+1級節點子集;所述步驟7:判斷所述資料對向量中是否還存在未被遍歷的資料對,若所述資料對向量中還存在未被遍歷的資料對,返回執行所述步驟1,繼續遍歷所述資料對向量中的下一個資料對;若所述資料對向量中的資料對被全部遍歷,結束遍歷。 The image detection method according to claim 2, wherein determining a plurality of node subsets according to the data pair vector includes: Step 1: starting from [a1, b1] to traverse the data pairs in the data pair vector; step 2: Determine whether the data pair satisfies the push-in stacking condition, if the data pair does not meet the push-in stacking condition, perform step 4; Step 3: if the data pair meets the push-in stacking condition, press the data pair into the stack; the step 4: judge whether the data pair meets the pop-up stacking condition, if the data pair does not meet the pop-up stacking condition, perform step 7; step 5: if the data pair meets the pop-up stacking condition, pop up the current The topmost data pair in the stack, and counting the number m of the remaining data pairs in the stack, where m represents a positive integer; Step 6: Determine the starting node corresponding to the topmost data pair in the stack, and determine that satisfies The data of the pop-up stacking condition corresponds to the end node, and the subset formed by all the nodes from the start node to the end node is set as the m+1 level node subset; the step 7: judge the data Check whether there are data pairs that have not been traversed in the vector, and if there are data pairs that have not been traversed in the data pair vector, return to step 1 and continue to traverse the next data pair in the data pair vector; If all the data pairs in the data pair vector have been traversed, the traversal ends. 如請求項2所述的圖像檢測方法,其中,所述根據所述多個節點子集確定所述神經網路模型中的運算元子組包括:遍歷所述多個節點子集;若所述多個節點子集中的任一個節點子集中的節點與其他節點子集中的節點都不重複,確定所述任一個節點子集為所述神經網路模型中的一個運算元子組。 The image detection method according to claim 2, wherein said determining the operator subset in the neural network model according to the plurality of node subsets includes: traversing the plurality of node subsets; if the The nodes in any one of the plurality of node subsets are not repeated with the nodes in other node subsets, and any one of the node subsets is determined as a subset of operands in the neural network model. 如請求項2所述的圖像檢測方法,其中,根據所述多個節點子集確定所述神經網路模型中的運算元子組還包括:遍歷所述多個節點子集,得到多個第一節點子集和多個第二節點子集,其中,所述第二節點子集真包含所述第一節點子集;去除所述第二節點子集中的所述第一節點子集的全部節點;設定所述第二節點子集中剩餘節點的集合為所述神經網路模型的運算元子組。 The image detection method according to claim 2, wherein, according to the multiple node subsets, determining the operator subset in the neural network model further includes: traversing the multiple node subsets to obtain multiple A first node subset and a plurality of second node subsets, wherein the second node subset really contains the first node subset; removing the first node subset from the second node subset All nodes; setting the set of remaining nodes in the second node subset as the operator subset of the neural network model. 如請求項5所述的圖像檢測方法,其中,根據所述多個節點子集確定所述神經網路模型中的運算元子組還包括:設定所述多個第一節點子集為所述神經網路模型的運算元子組。 The image detection method according to claim 5, wherein, according to the plurality of node subsets, determining the operator subset in the neural network model further includes: setting the plurality of first node subsets as all A subgroup of operands for a neural network model. 如請求項3所述的圖像檢測方法,其中:若資料對[ai,bi]中的bi大於等於2,確定所述資料對[ai,bi]滿足彈出堆疊條件;若資料對[ai,bi]中的ai大於等於2,確定所述資料對[ai,bi]滿足壓入堆疊條件。 The image detection method as described in claim 3, wherein: if the bi in the data pair [ai, bi] is greater than or equal to 2, it is determined that the data pair [ai, bi] meets the pop-up stacking condition; if the data pair [ai, bi] ai in bi] is greater than or equal to 2, it is determined that the data pair [ai, bi] satisfies the push-in stacking condition. 一種電子設備,其中,所述電子設備包括:記憶體,存儲至少一個指令;及處理器,執行所述記憶體中存儲的指令以實現如請求項1至請求項7中任意一項所述的圖像檢測方法。 An electronic device, wherein, the electronic device includes: a memory storing at least one instruction; and a processor executing the instruction stored in the memory to implement any one of claim 1 to claim 7. Image detection method. 一種電腦可讀存儲介質,其中:所述電腦可讀存儲介質中存儲有至少一個指令,所述至少一個指令被電子設備中的處理器執行以實現如請求項1至請求項7中任意一項所述的圖像檢測方法。 A computer-readable storage medium, wherein: at least one instruction is stored in the computer-readable storage medium, and the at least one instruction is executed by a processor in an electronic device to implement any one of request item 1 to request item 7 The image detection method described above.
TW110136617A 2021-09-30 2021-09-30 Method for detecting images, electronic device and storage medium TWI805005B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW110136617A TWI805005B (en) 2021-09-30 2021-09-30 Method for detecting images, electronic device and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW110136617A TWI805005B (en) 2021-09-30 2021-09-30 Method for detecting images, electronic device and storage medium

Publications (2)

Publication Number Publication Date
TW202316319A TW202316319A (en) 2023-04-16
TWI805005B true TWI805005B (en) 2023-06-11

Family

ID=86943208

Family Applications (1)

Application Number Title Priority Date Filing Date
TW110136617A TWI805005B (en) 2021-09-30 2021-09-30 Method for detecting images, electronic device and storage medium

Country Status (1)

Country Link
TW (1) TWI805005B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050059876A1 (en) * 2003-06-25 2005-03-17 Sriram Krishnan Systems and methods for providing automated regional myocardial assessment for cardiac imaging
TW201835784A (en) * 2016-12-30 2018-10-01 美商英特爾公司 The internet of things
CN110352432A (en) * 2017-04-07 2019-10-18 英特尔公司 Using for the improved training of deep neural network and the method and system of study
US20210201078A1 (en) * 2017-04-07 2021-07-01 Intel Corporation Methods and systems for advanced and augmented training of deep neural networks using synthetic data and innovative generative networks

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050059876A1 (en) * 2003-06-25 2005-03-17 Sriram Krishnan Systems and methods for providing automated regional myocardial assessment for cardiac imaging
TW201835784A (en) * 2016-12-30 2018-10-01 美商英特爾公司 The internet of things
CN110352432A (en) * 2017-04-07 2019-10-18 英特尔公司 Using for the improved training of deep neural network and the method and system of study
US20210201078A1 (en) * 2017-04-07 2021-07-01 Intel Corporation Methods and systems for advanced and augmented training of deep neural networks using synthetic data and innovative generative networks

Also Published As

Publication number Publication date
TW202316319A (en) 2023-04-16

Similar Documents

Publication Publication Date Title
AU2019340314B2 (en) Dynamic application migration between cloud providers
US7941420B2 (en) Method for organizing structurally similar web pages from a web site
CN111414987A (en) Training method and training device for neural network and electronic equipment
EP3161610B1 (en) Optimized browser rendering process
WO2022156180A1 (en) Similar text determination method and related device
WO2021196825A1 (en) Abstract generation method and apparatus, and electronic device and medium
WO2021051624A1 (en) Data acquisition method and apparatus, and electronic device and storage medium
US20200394448A1 (en) Methods for more effectively moderating one or more images and devices thereof
WO2020057523A1 (en) Method and device for triggering vulnerability detection
CN116861431B (en) Malicious software classification method and system based on multichannel image and neural network
US11436291B1 (en) Source rank metric of measuring sources of influence
CN106354587A (en) Mirror image server and method for exporting mirror image files of virtual machine
CN110502472A (en) A kind of the cloud storage optimization method and its system of large amount of small documents
TWI805005B (en) Method for detecting images, electronic device and storage medium
US20170124219A1 (en) Determining data field offsets using a document object model representation
CN117271645A (en) Test data processing method and device and computer readable storage medium
CN115002211B (en) Method, device, equipment and medium for realizing after-sale micro-service based on cloud protogenesis
US20170124156A1 (en) Search system
TWI777319B (en) Method and device for determining stem cell density, computer device and storage medium
CN112396151B (en) Rumor event analysis method, rumor event analysis device, rumor event analysis equipment and computer readable storage medium
US11055491B2 (en) Geographic location specific models for information extraction and knowledge discovery
CN110032445A (en) Big data Aggregation computation method and device
CN115906989A (en) Image detection method, electronic device, and storage medium
JP6351545B2 (en) Parameter management apparatus and parameter management program
CN118157970B (en) Suricata rule similarity calculation method based on editing distance