diff --git a/docs/assets/img/related_documents/v-training/3.jpg b/docs/assets/img/related_documents/v-training/3.jpg index d9cbf278..cf052b60 100644 Binary files a/docs/assets/img/related_documents/v-training/3.jpg and b/docs/assets/img/related_documents/v-training/3.jpg differ diff --git a/docs/en/core/m5stickv.md b/docs/en/core/m5stickv.md index 1e81d80c..50230238 100644 --- a/docs/en/core/m5stickv.md +++ b/docs/en/core/m5stickv.md @@ -191,7 +191,7 @@ M5StickV does not currently recognize all types of SD cards. We have tested some ### FUNCTIONAL DESCRIPTION #### 1.1 KENDRYTE K210 -The Kendryte K210 is a system-on-chip (SoC) that integrates machine vision and machine hearing. Using TSMC’s ultra-low-power 28-nm advanced process with dualcore 64-bit processors for better power efficiency, stability and reliability. The SoC strives for ”zero threshold” development and to be deployable in the user’s products in the shortest possible time, giving the product artificial intelligence

+The Kendryte K210 is a system-on-chip (SoC) that integrates machine vision. Using TSMC’s ultra-low-power 28-nm advanced process with dualcore 64-bit processors for better power efficiency, stability and reliability. The SoC strives for ”zero threshold” development and to be deployable in the user’s products in the shortest possible time, giving the product artificial intelligence

- Machine Vision - Better low power vision processing speed and accuracy - KPU high performance Convolutional Neural Network (CNN) hardware accelerator diff --git a/docs/en/related_documents/v-training.md b/docs/en/related_documents/v-training.md index 15db9e2b..59ddcab9 100644 --- a/docs/en/related_documents/v-training.md +++ b/docs/en/related_documents/v-training.md @@ -12,7 +12,9 @@ **[5. Run Recognition Program](#Run-Recognition-Program)** -## EasyLoader +

Users who have already programmed the firmware should start directly from the third step.

+ +## EasyLoader optional @@ -22,7 +24,7 @@ >2, After downloaded , double click to run the app, connect the device to computer via USB cable, select the com port number, then click "Burn" to start it. -## Download Firmware +## Download Firmware optional > EasyLoader is only Window-supported. If you don't have a Windows computer or you would like to download specific file to flash , please use "Kflash, download firmware below " @@ -56,7 +58,7 @@ > Material Training requires SD cards, users could downloade boot code zip files, unzip the files to SD card.(M5StickC only recognized certain type of SD card , [click to see the supported type](en/core/m5stickv?id=sd-card-test)) - + @@ -76,7 +78,7 @@ -note：In order to reach a certain accuracy, each Class should contains at least 120 pictures, or the Could Training would give out a rejection +note：In order to reach a certain accuracy, each Class should contains at least 35 pictures, or the Could Training would give out a rejection ## Material Checking and Compress @@ -87,7 +89,7 @@ > Inside folder "train","vaild", they share exact the same folder directory, when we switch Class, the program will generate the same folder (with a name of Class number) in both "train" and "vaild". The phtotos will placed either in "train" or "vaild", underneath the coorespondent Class folder. > Before we compress the package, we should check the photo and photo number, make sure for the same Class, the number of photos in the coorespondent Class Folder in -"train" and "vaild" should add up over 120. (like n1-n100 in train, n100-n120 in vaild). If any Class photos total amount were under 120, please either delete it or copy for back up. After finish the checking, let's compress the "train" and "vaild" to ZIP. +"train" and "vaild" should add up over 35. If any Class photos total amount were under 35, please either delete it or copy for back up. After finish the checking, let's compress the "train" and "vaild" to ZIP. ## Upload Data to Cloud diff --git a/docs/zh_CN/core/m5stickv.md b/docs/zh_CN/core/m5stickv.md index 93e48b02..6bd7b5b0 100644 --- a/docs/zh_CN/core/m5stickv.md +++ b/docs/zh_CN/core/m5stickv.md @@ -193,7 +193,7 @@ M5StickV目前并不能识别所有类型的SD卡，我们对一些常见的SD ### 功能描述 #### 1.1 KENDRYTE K210 -Kendryte K210 是集成机器视觉与机器听觉能力的系统级芯片 (SoC)。使用台积电 (TSMC) 超低功耗的 28 纳米先进制程，具有双核 64 位处理器，拥有较好的功耗性能，稳定性与可靠性。该方案力求零门槛开发，可在最短时效部署于用户的产品中，赋予产品人工智能.

+Kendryte K210 是集成机器视觉能力的系统级芯片 (SoC)。使用台积电 (TSMC) 超低功耗的 28 纳米先进制程，具有双核 64 位处理器，拥有较好的功耗性能，稳定性与可靠性。该方案力求零门槛开发，可在最短时效部署于用户的产品中，赋予产品人工智能.

- 具备机器视觉能力 - 更好的低功耗视觉处理速度与准确率 - 具备卷积人工神经网络硬件加速器 KPU，可高性能进行卷积人工神经网络运算 diff --git a/docs/zh_CN/related_documents/v-training.md b/docs/zh_CN/related_documents/v-training.md index d56ea51d..9423e183 100644 --- a/docs/zh_CN/related_documents/v-training.md +++ b/docs/zh_CN/related_documents/v-training.md @@ -12,7 +12,9 @@ **[5. 运行识别程序](#运行识别程序)** -## EasyLoader +

已经烧录了固件程序的用户请直接从第三步开始

+ +## EasyLoader 可选 @@ -22,7 +24,7 @@ >2.下载软件后，双击运行应用程序，将M5设备通过数据线连接至电脑,选择端口参数，点击 **"Burn"** 即可开始烧录 -## 下载固件 +## 下载固件 可选 > 需要指定烧录文件的用户可以选用**Kflash**进行固件烧录. @@ -49,7 +51,6 @@ >3.对于习惯使用命令行操作的用户来说还可以选择Kflash作为固件烧录工具.[点击此处查看详情](https://github.com/kendryte/kflash.py) - ## 训练素材拍摄 ### boot程序 @@ -57,7 +58,7 @@ > 拍摄训练素材需要使用到SD卡，用户需下载boot程序压缩包，并将压缩包内的所有文件解压放置到SD卡中（M5StickV对SD卡的选型有所要求，[点击此处查看支持类型](zh_CN/core/m5stickv?id=sd卡测试)） - + @@ -78,7 +79,7 @@ -注意：为了保证识别的准确率，每组Class拍摄素材张数需要超过120张，否则在进行云端训练时将不给予通过. +注意：为了保证识别的准确率，每组Class拍摄素材张数需要超过35张，否则在进行云端训练时将不给予通过. 素材的数量越多，识别训练的效果越好，识别率越高 ## 素材检查与压制 @@ -86,9 +87,9 @@ ->"train"、"vaild"两个文件夹中的Class序号文件夹目录是保持一致的，当切换Class并拍摄素材时，程序将会在"train"、"vaild"中同时创建Class序号一致的文件夹，并按照存放规则将所拍摄的图片分别存储到"train"、"vaild"各自目录下的Class文件夹中.（"train"文件夹中将存放拍摄的1 ~ 100、121 ~ n号素材图片."vaild"文件夹中将存放拍摄的101 ~ 120号素材图片.） +>"train"、"vaild"两个文件夹中的Class序号文件夹目录是保持一致的，当切换Class并拍摄素材时，程序将会在"train"、"vaild"中同时创建Class序号一致的文件夹，并按照存放规则将所拍摄的图片分别存储到"train"、"vaild"各自目录下的Class文件夹中. ->在压制打包前除了检查图片内容的正确性以外，必须确保"train"、"vaild"两个文件夹中同一Class序号目录里的素材图片总和大于120.数量总和小于120时的Class序号目录请自行删除或是备份处理.完成了检查工作，接下来要做就是素材文件的压制.将"train"、"vaild"两个文件夹通过压制工具压制为"zip"格式的压缩包. +>在压制打包前除了检查图片内容的正确性以外，必须确保"train"、"vaild"两个文件夹中同一Class序号目录里的素材图片总和大于35.数量总和小于35时的Class序号目录请自行删除或是备份处理.完成了检查工作，接下来要做就是素材文件的压制.将"train"、"vaild"两个文件夹通过压制工具压制为"zip"格式的压缩包. ## 数据上传云端