文件列表:

.idea/ (0, 2020-11-16)
.idea/.name (20, 2020-11-16)
.idea/codeStyles/ (0, 2020-11-16)
.idea/codeStyles/Project.xml (3567, 2020-11-16)
.idea/codeStyles/codeStyleConfig.xml (142, 2020-11-16)
.idea/gradle.xml (803, 2020-11-16)
.idea/jarRepositories.xml (1267, 2020-11-16)
.idea/misc.xml (357, 2020-11-16)
.idea/runConfigurations.xml (564, 2020-11-16)
.idea/vcs.xml (167, 2020-11-16)
amplify.json (145, 2020-11-16)
amplify/ (0, 2020-11-16)
amplify/.config/ (0, 2020-11-16)
amplify/.config/project-config.json (207, 2020-11-16)
amplify/backend/ (0, 2020-11-16)
amplify/backend/auth/ (0, 2020-11-16)
amplify/backend/auth/rockpaperscissor345eab27/ (0, 2020-11-16)
amplify/backend/auth/rockpaperscissor345eab27/parameters.json (1537, 2020-11-16)
amplify/backend/auth/rockpaperscissor345eab27/rockpaperscissor345eab27-cloudformation-template.yml (10868, 2020-11-16)
amplify/backend/backend-config.json (293, 2020-11-16)
amplify/backend/storage/ (0, 2020-11-16)
amplify/backend/storage/myStorage/ (0, 2020-11-16)
amplify/backend/storage/myStorage/parameters.json (1202, 2020-11-16)
amplify/backend/storage/myStorage/s3-cloudformation-template.json (12435, 2020-11-16)
amplify/backend/storage/myStorage/storage-params.json (2, 2020-11-16)
amplify/backend/tags.json (135, 2020-11-16)
amplify/team-provider-info.json (938, 2020-11-16)
app/ (0, 2020-11-16)
app/build.gradle (1660, 2020-11-16)
app/proguard-rules.pro (751, 2020-11-16)
app/src/ (0, 2020-11-16)
app/src/androidTest/ (0, 2020-11-16)
app/src/androidTest/java/ (0, 2020-11-16)
app/src/androidTest/java/com/ (0, 2020-11-16)
app/src/androidTest/java/com/rockpaperscissorsapp/ (0, 2020-11-16)
app/src/androidTest/java/com/rockpaperscissorsapp/ExampleInstrumentedTest.kt (676, 2020-11-16)
app/src/main/ (0, 2020-11-16)
... ...

# Rock-Paper-Scissor-App This repository contains my Rock Paper Scissor Classification App Project. It has been programmed using Kotlin and it uses AWS Sagemaker as a backend service to perform inference. The classification network is base on [MobileNet V2](https://arxiv.org/pdf/1801.04381.pdf) and fine tuned with manually recorded images using Keras. # General Idea and Motivation behind the Project The general motivation behind this project was, to learn more about different AWS Services. In addition, I wanted to try out, writing a new App using Kotlin. To this point in time, I only developed apps in native Java code (eg. [SpotFoxx](https://github.com/Jensssen/SpotFoxx)). Since I am a Machine Learning Engineer and ML is awesome, I decided to build an image based classification app. However, the ML part of this project was keeped as minimal as possible, so that I have more time to focus on the AWS backend part and a little bit on the Kotlin app. # Tech stack The following list shows most of the different programming languages, frameworks and tools that have been used for this project. - Kotlin -> Used to programm the app - Python -> Used inside of the Lambda Function, Sagemaker and NN training - Keras -> Used for the NN training - Git -> Version Control - AWS Amplify -> Great help to develop the App backend - AWS Services - IAM -> Identity management inside of AWS - Cognito -> App user identity management - S3 -> Storage of training data and NN Model - Lambda -> Used to access Sagemaker Endpoint - Sagemaker -> Used to host endpoint that performs inference # Final Result The following Link leads to a YouTube Video that demonstrates the App. [VIDEO](https://youtu.be/dJI1shZAKDk) # Architecture The following image show the underlying project architecture. A small remark: The architecture has not been developed in a way that it is very cost efficient. For example the kotlin app could call the Sagemaker Endpoint directly and no lambda function is needed in between. Instead, I wanted to try out many different things eg. how lambda functions work. Therefore, I implemented it like this.  # Machine Learning Related Code The ML related code can be found in an extra repository which can be found [here](https://github.com/Jensssen/rock_paper_scissor_classification). As already mentioned, the main focus of this project was not ML. Therefore, the ML part is keeped very minimal.

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：