TWI717043B - System and method for recognizing aquatic creature - Google Patents

System and method for recognizing aquatic creature Download PDF

Info

Publication number
TWI717043B
TWI717043B TW108135618A TW108135618A TWI717043B TW I717043 B TWI717043 B TW I717043B TW 108135618 A TW108135618 A TW 108135618A TW 108135618 A TW108135618 A TW 108135618A TW I717043 B TWI717043 B TW I717043B
Authority
TW
Taiwan
Prior art keywords
underwater
creature
neural network
ultrasonic
image
Prior art date
Application number
TW108135618A
Other languages
Chinese (zh)
Other versions
TW202115425A (en
Inventor
黃建國
沈煒翔
Original Assignee
佳世達科技股份有限公司
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 佳世達科技股份有限公司 filed Critical 佳世達科技股份有限公司
Priority to TW108135618A priority Critical patent/TWI717043B/en
Application granted granted Critical
Publication of TWI717043B publication Critical patent/TWI717043B/en
Publication of TW202115425A publication Critical patent/TW202115425A/en

Links

Images

Landscapes

  • Image Analysis (AREA)

Abstract

A system for recognizing an aquatic creature includes an underwater ultrasonic device, a processor and a mixed neural network. The underwater ultrasonic device transmits a plurality of ultrasonic signals to the aquatic creature and receives a plurality of reflected signals of the ultrasonic signals. The processor generates an ultrasonic image according to the reflected signals. The mixed neural network includes a plurality of self-training features and a plurality of manual setting features associated with the aquatic creature. The mixed neural network recognizes the aquatic creature from the ultrasonic image according to the self-training features and the manual setting features.

Description

水下生物辨識系統及水下生物辨識方法 Underwater biological identification system and underwater biological identification method

本發明關於一種水下生物辨識系統及水下生物辨識方法，尤指一種可幫助使用者快速地自超音波影像辨識出水下生物之水下生物辨識系統及水下生物辨識方法。 The present invention relates to an underwater biological identification system and an underwater biological identification method, in particular to an underwater biological identification system and an underwater biological identification method that can help a user to quickly identify underwater creatures from ultrasonic images.

由於超音波具有不破壞材料結構以及不傷害生物體的特性，因而普遍地被應用於水下生物的偵測。超音波可以讓使用者觀察水下生物而不需直接潛入水中。然而，沒有受過專業訓練的一般使用者並無法直接從超音波影像辨識出水下生物的種類。因此，如何幫助使用者快速地自超音波影像辨識出水下生物，便成為一個重要的研究課題。 Since ultrasound has the characteristics of not destroying the material structure and not harming the organism, it is widely used in the detection of underwater creatures. Ultrasonic waves allow users to observe underwater creatures without diving directly into the water. However, ordinary users without professional training cannot directly identify the types of underwater creatures from ultrasonic images. Therefore, how to help users quickly identify underwater creatures from ultrasound images has become an important research topic.

本發明的目的之一在於提供一種可幫助使用者快速地自超音波影像辨識出水下生物之水下生物辨識系統及水下生物辨識方法，以解決上述問題。 One of the objectives of the present invention is to provide an underwater biological identification system and underwater biological identification method that can help users quickly identify underwater creatures from ultrasonic images, so as to solve the above-mentioned problems.

根據一實施例，本發明之水下生物辨識系統包含一水下超音波裝置、一處理器以及一混合型神經網路。水下超音波裝置朝一水下生物發射複數個超音波訊號，且接收複數個超音波訊號之複數個反射訊號。處理器根據複數個反射訊號產生一超音波影像。混合型神經網路包含關於水下生物之複數個自我訓練特徵以及複數個人工設定特徵。混合型神經網路根據複數個自我訓練特徵與複數個人工設定特徵，自超音波影像辨識出水下生物。 According to one embodiment, the underwater biological recognition system of the present invention includes an underwater ultrasonic device, a processor, and a hybrid neural network. The underwater ultrasonic device transmits a plurality of ultrasonic signals to an underwater creature, and receives a plurality of reflection signals of the plurality of ultrasonic signals. The processor generates an ultrasonic image based on a plurality of reflected signals. The hybrid neural network contains a plurality of self-training features and a plurality of artificial setting features about underwater creatures. The hybrid neural network recognizes underwater creatures from ultrasonic images based on multiple self-training features and multiple artificially set features.

根據另一實施例，本發明之水下生物辨識方法包含下列步驟：以一 水下超音波裝置朝一水下生物發射複數個超音波訊號，且接收複數個超音波訊號之複數個反射訊號；根據複數個反射訊號產生一超音波影像；將超音波影像輸入一混合型神經網路，其中混合型神經網路包含關於水下生物之複數個自我訓練特徵以及複數個人工設定特徵；以及由混合型神經網路根據複數個自我訓練特徵與複數個人工設定特徵，自超音波影像辨識出水下生物。 According to another embodiment, the underwater biological identification method of the present invention includes the following steps: The underwater ultrasonic device emits a plurality of ultrasonic signals to an underwater creature, and receives a plurality of reflection signals of the plurality of ultrasonic signals; generates an ultrasonic image based on the plurality of reflection signals; inputs the ultrasonic image into a hybrid neural network Road, where the hybrid neural network includes multiple self-training features and multiple artificial setting features about underwater creatures; and the hybrid neural network is based on the multiple self-training features and multiple artificial setting features from ultrasound images Identify underwater creatures.

綜上所述，本發明係利用混合型神經網路對超音波影像進行分析，以自超音波影像辨識出水下生物。由於混合型神經網路包含關於水下生物之複數個自我訓練特徵以及複數個人工設定特徵，因此，本發明可精準地對超音波影像中的水下生物進行辨識與分類，進而幫助使用者快速地自超音波影像辨識出水下生物。 In summary, the present invention uses a hybrid neural network to analyze ultrasonic images to identify underwater creatures from the ultrasonic images. Since the hybrid neural network includes a plurality of self-training features and a plurality of artificial setting features about underwater creatures, the present invention can accurately identify and classify underwater creatures in ultrasonic images, thereby helping users quickly The ground has identified underwater creatures from ultrasonic images.

關於本發明之優點與精神可以藉由以下的發明詳述及所附圖式得到進一步的瞭解。 The advantages and spirit of the present invention can be further understood from the following detailed description of the invention and the accompanying drawings.

1:水下生物辨識系統 1: Underwater biometric system

3:超音波影像 3: Ultrasonic image

10:水下超音波裝置 10: Underwater ultrasonic device

12:處理器 12: processor

14:混合型神經網路 14: Hybrid neural network

16:資料庫 16: Database

30:水下生物 30: Underwater creatures

32:標籤 32: label

S10-S16:步驟 S10-S16: steps

第1圖為根據本發明一實施例之水下生物辨識系統的功能方塊圖。 Figure 1 is a functional block diagram of an underwater biometric system according to an embodiment of the invention.

第2圖為根據本發明一實施例之超音波影像的示意圖。 FIG. 2 is a schematic diagram of an ultrasonic image according to an embodiment of the invention.

第3圖為根據本發明一實施例之水下生物辨識方法的流程圖。 Figure 3 is a flowchart of an underwater creature identification method according to an embodiment of the invention.

第4圖為第2圖中的水下生物之目前影像轉換為水下生物之實際影像的示意圖。 Figure 4 is a schematic diagram of the current image of underwater creatures in Figure 2 transformed into actual images of underwater creatures.

第5圖為以標籤標示水下生物的示意圖。 Figure 5 is a schematic diagram of labeling underwater creatures.

請參閱第1圖至第3圖，第1圖為根據本發明一實施例之水下生物辨識系統1的功能方塊圖，第2圖為根據本發明一實施例之超音波影像3的示意圖，第3圖為根據本發明一實施例之水下生物辨識方法的流程圖。第3圖中的水下生物辨 識方法可以第1圖中的水下生物辨識系統1來實現。 Please refer to Figures 1 to 3. Figure 1 is a functional block diagram of an underwater biometric identification system 1 according to an embodiment of the present invention, and Figure 2 is a schematic diagram of an ultrasonic image 3 according to an embodiment of the present invention. Figure 3 is a flowchart of an underwater creature identification method according to an embodiment of the invention. Underwater bioassay in image 3 The recognition method can be implemented by the underwater biological recognition system 1 in Figure 1.

如第1圖所示，水下生物辨識系統1包含一水下超音波裝置10、一處理器12、一混合型神經網路14以及一資料庫16。於此實施例中，處理器12、混合型神經網路14與資料庫16可設置於電腦(未顯示)中，且電腦可與水下超音波裝置10形成通訊，以進行訊號傳輸。水下超音波裝置10可為一水下超音波探頭或其它可收發超音波之水下超音波裝置。 As shown in FIG. 1, the underwater biometric identification system 1 includes an underwater ultrasonic device 10, a processor 12, a hybrid neural network 14 and a database 16. In this embodiment, the processor 12, the hybrid neural network 14 and the database 16 can be set in a computer (not shown), and the computer can communicate with the underwater ultrasonic device 10 for signal transmission. The underwater ultrasonic device 10 may be an underwater ultrasonic probe or other underwater ultrasonic devices capable of transmitting and receiving ultrasonic waves.

當使用者欲以水下生物辨識系統1對一水下生物進行辨識時，使用者可先以水下超音波裝置10朝水下生物發射複數個超音波訊號，且接收複數個超音波訊號之複數個反射訊號(第3圖中的步驟S10)。接著，處理器12即會根據複數個反射訊號產生如第2圖所示之超音波影像3(第3圖中的步驟S12)，其中超音波影像3中存在水下生物30(例如，魚、蝦、蟹等)。於此實施例中，超音波影像3可為經二值化處理後之二值化超音波影像，但不以此為限。需說明的是，二值化處理技術係為習知技藝之人所熟知，在此不再贅述。此外，超音波影像3中的水下生物之數量與種類係根據實際應用而決定，不以圖中所繪示之實施例為限。 When the user wants to identify an underwater creature with the underwater biological recognition system 1, the user can first use the underwater ultrasonic device 10 to transmit a plurality of ultrasonic signals to the underwater creature, and receive one of the plurality of ultrasonic signals Multiple reflection signals (Step S10 in Figure 3). Then, the processor 12 generates an ultrasonic image 3 as shown in Figure 2 according to the plurality of reflection signals (step S12 in Figure 3), in which the ultrasonic image 3 contains underwater creatures 30 (for example, fish, Shrimp, crab, etc.). In this embodiment, the ultrasonic image 3 may be a binary ultrasonic image after binarization processing, but it is not limited to this. It should be noted that the binarization processing technology is well-known to those who are familiar with the art, and will not be repeated here. In addition, the number and types of underwater creatures in the ultrasonic image 3 are determined according to actual applications, and are not limited to the embodiment shown in the figure.

接著，處理器12可將超音波影像3輸入混合型神經網路14(第3圖中的步驟S14)，以對超音波影像3中的水下生物30進行辨識與分類。本發明之混合型神經網路14包含關於水下生物30之複數個自我訓練特徵以及複數個人工設定特徵。於此實施例中，混合型神經網路14可為卷積神經網路(Convolution Neural Network，CNN)或其它類似神經網路。於此實施例中，混合型神經網路14係已預先被訓練好，用以辨識水下生物。混合型神經網路14可預先對包含水下生物30之一影片進行影像辨識，以得到複數個自我訓練特徵。上述之影片可包含水下生物30之位置資訊、運動資訊、器官資訊等，視實際應用而定。需說明的是，神經網路之詳細訓練過程係為習知技藝之人所熟知，在此不再贅述。 Then, the processor 12 can input the ultrasonic image 3 into the hybrid neural network 14 (step S14 in FIG. 3) to identify and classify the underwater creatures 30 in the ultrasonic image 3. The hybrid neural network 14 of the present invention includes a plurality of self-training features and a plurality of artificial setting features about underwater creatures 30. In this embodiment, the hybrid neural network 14 may be a Convolution Neural Network (CNN) or other similar neural networks. In this embodiment, the hybrid neural network 14 has been pre-trained to recognize underwater creatures. The hybrid neural network 14 can perform image recognition on a video containing underwater creatures 30 in advance to obtain a plurality of self-training features. The above-mentioned video may include position information, movement information, organ information, etc. of the underwater creature 30, depending on the actual application. It should be noted that the detailed training process of the neural network is well-known to those who have learned the skills, and will not be repeated here.

此外，複數個人工設定特徵可以手動方式輸入混合型神經網路14。 舉例而言，當水下生物為魚類時，人工設定特徵可包含魚體長度、魚體大小、游泳速度、魚鰾大小、活動深度、雷納數(Reynolds number)等，視實際應用而定。當使用者在超音波影像上發現新水下生物時，使用者可以自行計算並輸入上述之人工設定特徵至混合型神經網路14，以增加混合型神經網路14辨識水下生物的準確率。 In addition, a plurality of artificially set characteristics can be manually input into the hybrid neural network 14. For example, when the underwater creature is a fish, the artificially set features may include fish body length, fish body size, swimming speed, swim bladder size, activity depth, Reynolds number, etc., depending on the actual application. When the user finds a new underwater creature on the ultrasound image, the user can calculate and input the above artificially set features into the hybrid neural network 14 to increase the accuracy of the hybrid neural network 14 in identifying underwater creatures .

因此，在超音波影像3輸入混合型神經網路14後，即可由混合型神經網路14根據複數個自我訓練特徵與複數個人工設定特徵，自超音波影像3辨識出水下生物30(第3圖中的步驟S16)。 Therefore, after the ultrasonic image 3 is input to the hybrid neural network 14, the hybrid neural network 14 can identify underwater creatures 30 from the ultrasonic image 3 based on a plurality of self-training features and a plurality of artificially set features (third Step S16 in the figure).

請參閱第4圖，第4圖為第2圖中的水下生物30之目前影像轉換為水下生物30之實際影像的示意圖。於此實施例中，本發明可進一步提供水下生物30之一實際影像，其中資料庫16可用以儲存水下生物30之實際影像。在混合型神經網路14自超音波影像3辨識出水下生物30後，處理器12可將超音波影像3中水下生物30之目前影像(如第2圖所示)轉換為水下生物30之實際影像(如第4圖所示)。藉此，使用者即可藉由水下生物30之實際影像快速地自超音波影像3辨識出水下生物30。 Please refer to Fig. 4, which is a schematic diagram of the current image of the underwater creature 30 in Fig. 2 being converted to the actual image of the underwater creature 30. In this embodiment, the present invention can further provide an actual image of the underwater creature 30, wherein the database 16 can be used to store the actual image of the underwater creature 30. After the hybrid neural network 14 has identified the underwater creature 30 from the ultrasonic image 3, the processor 12 can convert the current image of the underwater creature 30 in the ultrasonic image 3 (as shown in Figure 2) into the underwater creature 30 The actual image (as shown in Figure 4). In this way, the user can quickly recognize the underwater creature 30 from the ultrasonic image 3 based on the actual image of the underwater creature 30.

請參閱第5圖，第5圖為以標籤32標示水下生物30的示意圖。於此實施例中，本發明可進一步提供水下生物30之一標籤32(例如，魚類名稱)，其中資料庫16可用以儲存水下生物30之標籤32。在混合型神經網路14自超音波影像3辨識出水下生物30後，處理器12可以標籤32標示水下生物30，使得使用者可藉由水下生物30之標籤32快速地自超音波影像3辨識出水下生物30。如第5圖所示，標籤32可標示於水下生物30之實際影像周圍。當然，標籤32亦可標示於水下生物30之實際影像上，視實際應用而定。此外，標籤32亦可標示於如第2圖所示之水下生物30之目前影像上或周圍，視實際應用而定。 Please refer to FIG. 5, which is a schematic diagram of marking underwater creatures 30 with tags 32. In this embodiment, the present invention can further provide a tag 32 (for example, a fish name) of the underwater creature 30, wherein the database 16 can be used to store the tag 32 of the underwater creature 30. After the hybrid neural network 14 recognizes the underwater creature 30 from the ultrasonic image 3, the processor 12 can label the underwater creature 30 with the tag 32, so that the user can quickly read the ultrasonic image from the tag 32 of the underwater creature 30 3 Identify underwater creatures 30. As shown in FIG. 5, the tag 32 can be marked around the actual image of the underwater creature 30. Of course, the tag 32 can also be marked on the actual image of the underwater creature 30, depending on the actual application. In addition, the tag 32 can also be marked on or around the current image of the underwater creature 30 as shown in Figure 2, depending on the actual application.

綜上所述，本發明係利用混合型神經網路對超音波影像進行分析， 以自超音波影像辨識出水下生物。由於混合型神經網路包含關於水下生物之複數個自我訓練特徵以及複數個人工設定特徵，因此，本發明可精準地對超音波影像中的水下生物進行辨識與分類，進而幫助使用者快速地自超音波影像辨識出水下生物。此外，本發明可進一步提供水下生物之實際影像及/或標籤，使得使用者可藉由水下生物之實際影像及/或標籤快速地自超音波影像辨識出水下生物。 In summary, the present invention uses a hybrid neural network to analyze ultrasound images, Identify underwater creatures with self-ultrasonic images. Since the hybrid neural network includes a plurality of self-training features and a plurality of artificial setting features about underwater creatures, the present invention can accurately identify and classify underwater creatures in ultrasonic images, thereby helping users quickly The ground has identified underwater creatures from ultrasonic images. In addition, the present invention can further provide actual images and/or tags of underwater creatures, so that users can quickly recognize underwater creatures from ultrasonic images by using the actual images and/or tags of underwater creatures.

以上所述僅為本發明之較佳實施例，凡依本發明申請專利範圍所做之均等變化與修飾，皆應屬本發明之涵蓋範圍。 The foregoing descriptions are only preferred embodiments of the present invention, and all equivalent changes and modifications made in accordance with the scope of the patent application of the present invention shall fall within the scope of the present invention.

1:水下生物辨識系統 1: Underwater biometric system

10:水下超音波裝置 10: Underwater ultrasonic device

12:處理器 12: processor

14:混合型神經網路 14: Hybrid neural network

16:資料庫 16: Database

Claims (8)

一種水下生物辨識系統，包含：一水下超音波裝置，朝一水下生物發射複數個超音波訊號，且接收該複數個超音波訊號之複數個反射訊號；一處理器，根據該複數個反射訊號產生一超音波影像；一混合型神經網路，包含關於該水下生物之複數個自我訓練特徵以及複數個人工設定特徵，該混合型神經網路根據該複數個自我訓練特徵與該複數個人工設定特徵，自該超音波影像辨識出該水下生物；以及一資料庫，儲存該水下生物之一實際影像；其中，在該混合型神經網路自該超音波影像辨識出該水下生物後，該處理器將該超音波影像中該水下生物之一目前影像轉換為該水下生物之該實際影像。 An underwater creature identification system, comprising: an underwater ultrasonic device, which transmits a plurality of ultrasonic signals to an underwater creature, and receives a plurality of reflection signals of the plurality of ultrasonic signals; a processor, based on the plurality of reflections The signal generates an ultrasound image; a hybrid neural network including a plurality of self-training features and a plurality of artificially set features about the underwater creature, the hybrid neural network is based on the plurality of self-training features and the plurality of Artificially set features to identify the underwater creature from the ultrasound image; and a database storing an actual image of the underwater creature; wherein the hybrid neural network identifies the underwater creature from the ultrasound image After the creature, the processor converts the current image of one of the underwater creatures in the ultrasonic image into the actual image of the underwater creature. 如請求項1所述之水下生物辨識系統，其中該混合型神經網路預先對包含該水下生物之一影片進行影像辨識，以得到該複數個自我訓練特徵。 The underwater biological recognition system according to claim 1, wherein the hybrid neural network performs image recognition on a video containing the underwater biological in advance to obtain the plurality of self-training features. 如請求項1所述之水下生物辨識系統，其中該複數個人工設定特徵係以手動方式輸入該混合型神經網路。 The underwater biometric system according to claim 1, wherein the plurality of artificially set features are manually input into the hybrid neural network. 如請求項1所述之水下生物辨識系統，其中該資料庫儲存該水下生物之一標籤；在該混合型神經網路自該超音波影像辨識出該水下生物後，該處理器以該標籤標示該水下生物。 The underwater creature identification system of claim 1, wherein the database stores a tag of the underwater creature; after the hybrid neural network recognizes the underwater creature from the ultrasonic image, the processor uses The label indicates the underwater creature. 一種水下生物辨識方法，包含下列步驟：以一水下超音波裝置朝一水下生物發射複數個超音波訊號，且接收該複數個超音波訊號之複數個反射訊號；根據該複數個反射訊號產生一超音波影像；將該超音波影像輸入一混合型神經網路，其中該混合型神經網路包含關 於該水下生物之複數個自我訓練特徵以及複數個人工設定特徵；由該混合型神經網路根據該複數個自我訓練特徵與該複數個人工設定特徵，自該超音波影像辨識出該水下生物；提供該水下生物之一實際影像；以及在該混合型神經網路自該超音波影像辨識出該水下生物後，將該超音波影像中該水下生物之一目前影像轉換為該水下生物之該實際影像。 An underwater creature identification method, comprising the following steps: using an underwater ultrasonic device to transmit a plurality of ultrasonic signals to an underwater creature, and receiving a plurality of reflection signals of the plurality of ultrasonic signals; generating according to the plurality of reflection signals An ultrasound image; the ultrasound image is input into a hybrid neural network, wherein the hybrid neural network includes A plurality of self-training features and a plurality of artificially set characteristics of the underwater creature; the hybrid neural network recognizes the underwater from the ultrasonic image according to the plurality of self-training characteristics and the plurality of artificially set characteristics Creature; providing an actual image of the underwater creature; and after the hybrid neural network recognizes the underwater creature from the ultrasonic image, convert one of the current images of the underwater creature in the ultrasonic image into the This actual image of underwater creatures. 如請求項5所述之水下生物辨識方法，其中該混合型神經網路預先對包含該水下生物之一影片進行影像辨識，以得到該複數個自我訓練特徵。 The underwater creature identification method according to claim 5, wherein the hybrid neural network performs image recognition on a video containing the underwater creature in advance to obtain the plurality of self-training features. 如請求項5所述之水下生物辨識方法，其中該複數個人工設定特徵係以手動方式輸入該混合型神經網路。 The underwater biological identification method according to claim 5, wherein the plurality of artificially set features are manually input into the hybrid neural network. 如請求項5所述之水下生物辨識方法，另包含下列步驟：提供該水下生物之一標籤；以及在該混合型神經網路自該超音波影像辨識出該水下生物後，以該標籤標示該水下生物。 The underwater creature identification method according to claim 5, further comprising the following steps: providing a tag of the underwater creature; and after the hybrid neural network recognizes the underwater creature from the ultrasonic image, using the The label indicates the underwater creature.
TW108135618A 2019-10-02 2019-10-02 System and method for recognizing aquatic creature TWI717043B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW108135618A TWI717043B (en) 2019-10-02 2019-10-02 System and method for recognizing aquatic creature

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW108135618A TWI717043B (en) 2019-10-02 2019-10-02 System and method for recognizing aquatic creature

Publications (2)

Publication Number Publication Date
TWI717043B true TWI717043B (en) 2021-01-21
TW202115425A TW202115425A (en) 2021-04-16

Family

ID=75237552

Family Applications (1)

Application Number Title Priority Date Filing Date
TW108135618A TWI717043B (en) 2019-10-02 2019-10-02 System and method for recognizing aquatic creature

Country Status (1)

Country Link
TW (1) TWI717043B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TW571248B (en) * 2001-10-11 2004-01-11 Exscientia Llc Method and apparatus for learning to classify patterns and assess the value of decisions
GB2522302A (en) * 2013-10-31 2015-07-22 Furuno Electric Co Size-and-type determining device, underwater detecting apparatus and method of determining size and type
WO2016205938A1 (en) * 2015-06-22 2016-12-29 Appetite Lab Inc. Devices and methods for locating and visualizing underwater objects

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TW571248B (en) * 2001-10-11 2004-01-11 Exscientia Llc Method and apparatus for learning to classify patterns and assess the value of decisions
GB2522302A (en) * 2013-10-31 2015-07-22 Furuno Electric Co Size-and-type determining device, underwater detecting apparatus and method of determining size and type
WO2016205938A1 (en) * 2015-06-22 2016-12-29 Appetite Lab Inc. Devices and methods for locating and visualizing underwater objects

Also Published As

Publication number Publication date
TW202115425A (en) 2021-04-16

Similar Documents

Publication Publication Date Title
Miech et al. Learning a text-video embedding from incomplete and heterogeneous data
US11093560B2 (en) Stacked cross-modal matching
US11983950B2 (en) Entity identification using machine learning
CN109658938B (en) Method, device and equipment for matching voice and text and computer readable medium
TWI731397B (en) Method for verifying training data, training system, and computer program product
EP3933686A3 (en) Video processing method, apparatus, electronic device, storage medium, and program product
WO2020023467A1 (en) Unique identification of freely swimming fish in an aquaculture environment
CN108922622A (en) A kind of animal health monitoring method, device and computer readable storage medium
WO2021088935A1 (en) Adversarial network architecture optimization method and system, and image description generation method and system
CN107818314A (en) Face image processing method, device and server
WO2021120779A1 (en) Human-machine dialogue-based user profile construction method and system, and terminal and storage medium
WO2021135546A1 (en) Deep neural network interpretation method and device, terminal, and storage medium
CN111460121A (en) Visual semantic conversation method and system
CN111539937A (en) Object index detection method and livestock weight detection method and device
CN114239612A (en) Multi-modal neural machine translation method, computer equipment and storage medium
CN116778309A (en) Residual bait monitoring method, device, system and storage medium
Merz et al. Onset of melanophore patterns in the head region of chinook salmon: a natural marker for the reidentification of individual fish
TWI717043B (en) System and method for recognizing aquatic creature
Pearson et al. Testing and deployment of C-VISS (cetacean-borne video camera and integrated sensor system) on wild dolphins
Wang et al. A Real-Time Individual Identification Method for Swimming Fish Based on Improved Yolov5
Napier et al. Using mobile-based augmented reality and object detection for real-time Abalone growth monitoring
CN116578738A (en) Graph-text retrieval method and device based on graph attention and generating countermeasure network
CN110764093A (en) Underwater biological identification system and method thereof
Shapiro Recognition of individuals within the social group: signature vocalizations
Dash et al. Fish Type and Disease Classification Using Deep Learning Model Based Customized CNN with Resnet 50 Technique.