The community from sunbird-saral

Reduce Saral App bundle/APK size

Options to consider:
gradle.properties
android.enableR8=true

build.gradle
shrinkResources enableProguardInReleaseBuilds
zipAlignEnabled enableProguardInReleaseBuilds
useProguard enableProguardInReleaseBuilds
// resConfigs "en"
minifyEnabled enableProguardInReleaseBuilds
proguardFiles getDefaultProguardFile("proguard-android-optimize.txt"), "proguard-rules.pro"

Snapshot of current APK size:

As deployer/Maintainer maintain Unit Tests for Reference Backend

Identify NodeJS framework(eg. Jest) for Unit Tests
Build Unit Test Wrappers and package installations
Unit Tests Writing for APIs
Integrating with Jenkins pipelines

Multi-user support in reference backend as configured in u.p. Backend implementation

Locking the exams for scanning(recommencation from u.p implementation)

Locking the exams for scanning(recommendation from u.p implementation)

Reference backend implementation of locking functionality at the state/district/school level
App integration with locking functionality
Documentation

Tentative Dev Completion Date: 28th Nov 2022

Anganawadi beneficiary growth Monitoring Layout Design,ROI Generation,Testing

Upgrade deprecated packages and fix security vulnerabilities

Current status:

14 vulnerabilities (7 high, 7 critical)
npm WARN deprecated [email protected]
npm WARN deprecated [email protected]

Saral App Non-Black Pen,Pencil filled OMR Detection

Discussed in https://github.com/orgs/Sunbird-Saral/discussions/12

^{Originally posted by karsood August 19, 2022}
Observation: With students filling out the OMR form, bringing the control on choice of pen will require extreme admin control.

Desire: Enhanced accuracy for blue ink + other colours

To Support Non-Black Pen,Pencil filled OMR Detection as experimental feature that can be enabled with flag in layout roi json

As a user i should be able to review the scanned results for a student/layout

As a user, I should be able to review the scanned results for a student/layout.

On the Scan Status screen include a Review button
The review button should show up scan results stored on-device storage for review before pushing to the backend.
The user should be able to close the review screen to go back to Scan Status Screen.
In Minimal Mode, the Scan Data screen should have the same Review button to review scan results.
Tentative Dev Completion Date: 28th Nov.

Login API Auth token should have actual school or orgId as userId instead of mongo internal id

Login API Auth token should have actual school or orgId as userId instead of mongo internal id .

when we are skipping a student id then same student id on next page should not show error message "ID Shouldn't be duplicated"

when we are skipping a student id then same student id on next page should not show error message "ID Shouldn't be duplicated":

Tentative Dev Completion Date: 25th Nov 2022

Saral App global Cache clear functionality

Saral App global Cache clear functionality:

/brand/default API response is currently cached as a global cache in offlineMode. When the user clears the All User Cache, this default branding should be fetched from the backend API for the first time.

As deployer/Maintainer I should be able to improve Handwritten Digits model Accuracy by experimenting with open and synthetic datasets - Iteration 3

As deployer/Maintainer I should be able to improve Handwritten Digits Accuracy by experimenting with open and synthetic datasets - Iteration 3. Expected accuracy improvement is approximately 3% to 5%.

Experiment with the existing model version (https://github.com/Sunbird-Saral/Project-Saral/blob/main/v1.0/frontend/SaralApp/android/saralsdk/src/main/assets/trained_resnet_real_synthetic_v2_20.tflite) to identify the miss-classifications.
Identifying open data sets
Experiment with new data with the existing model and check the results and Prepare the new data
Training the next version of the model
Inferencing and identifying optimal model version/checkpoint.
Package with SaralSDK/SaralReference App to test in real-time.

As a deployer/Maintainer, I should be able to improve the Handwritten Block letter Alphanumeric model Accuracy with production datasets - Iteration 3

As a deployer/Maintainer, I should be able to improve the Handwritten Block letter Alphanumeric model Accuracy with production datasets - Iteration 3 :

Pre-requisite: Get handwritten layout sheets from the field with different users (considering demographics like gender, geography etc. of the users)

Scan and collect image data of miss-predicted Handwritten Block Letter Alphanumeric models from the layout sheets filled by various users.
Generate Synthetic data from production data as needed.
Prepare complete training data
Experiment with the existing model version (https://github.com/Sunbird-Saral/Project-Saral/tree/main/v1.0/frontend/SaralApp/android/saralsdk/src/main/assets#:~:text=model_letter_digits_finetune_18_07_22_epoch_15.tflite) to identify the miss-classifications.
Scan and
Training the next version of the model.
Inferencing and identifying optimal model version/checkpoint.
Package with SaralSDK/SaralReference App to test in real-time.

Block Letter English Alphabets Detection through Saral App

As a deployer/Maintainer, load test the reference backend APIs, fine-tune code and config, and publish the recommendations to scale

As a deployer/Maintainer, load test the reference backend APIs, fine-tune code and config, and publish the recommendations to scale.

Identify performance testing tools (example: https://www.artillery.io/, https://github.com/artilleryio/artillery )
Identify and configure Load Test scenarios
Configure app metrics tools for monitoring ( example: https://www.npmjs.com/package/appmetrics-dash)
Perform Load test with configured scenarios
Finetune the code and configurations if any bottlenecks and rerun load tests
Document configuration recommendations to scale the backend APIs.

Personal Details Layout generation , configuration , testing with Alphanumeric block letters support

Personal Details Layout generation , configuration, testing with Alphanumeric block letters support

As deployer/Maintainer experiment on a free English text extraction model that can be embedded into Saral SDK/Reference App.

Explore and Identify suitable ML model architectures for free English text extraction with the below characteristics.
- Model should be convertible to TFLite so that it can be embedded into Android App
- TFLite model size should be less than 20 MB to start with
- Free English Text extraction method is mainly expected to Phygitize fields like name, and address in a physical layout instead of character-by-character classification
- In SaralSDK/Reference app this model should be able to function offline using mobile computing power.

Use a custom TensorFlow Lite model on Android support - To be reviewed.

Use a custom TensorFlow Lite model on Android support using the below approach. This will add a new deployment model to upgrade to a new version of ML models dynamically without changing the app.

https://firebase.google.com/docs/ml/android/use-custom-models

As a deployer/Maintainer profile the Reference Backend Code and Finetune for production readiness

As a deployer/Maintainer profile the Reference Backend Code for production readiness

Identify NodeJS-based code profiler to identify code bottlenecks
Configure profiler
Identify Code bottlenecks and fix
Identify Packaging optimizations(Node packaging into Docker image)
Configuration optimization (Thread Pooling, Caching)

As deployer/Maintainer I should be able to improve the Handwritten Digits model Accuracy with production datasets - Iteration 4

As deployer/Maintainer I should be able to improve the Handwritten Digits model Accuracy with production datasets - Iteration 4 :

As deployer/Maintainer I should be able to improve Handwritten Digits Accuracy by experimenting with open and synthetic datasets - Iteration 4.

Pre-requisite: Get handwritten layout sheets from the field with different users (considering demographics like gender, geography etc. of the users)

Scan and collect image data of miss-predicted digits from the layout sheets filled by various users
Generate Synthetic data from production data as needed.
Prepare complete training data
Training the next version of the model
Inferencing and identifying optimal model version/checkpoint.
Package with SaralSDK/SaralReference App to test in real-time.

Saral App Icon configuration documentation

Document steps on how to replace the default saral app icon for a given implementation.
Once steps are reviewed, can be included in the below microsite link.

https://saral.sunbird.org/use/generate-aab-app-bundle-from-source-code

Saral App Size optimization

Conditional App building strategy based on the models/features to be supported.
Reduce size of assets
Clean up unwanted assets if any

As deployer/Maintainer profile reference backend code and finetune the code for production readiness

As deployer/Maintainer profile reference backend code and finetune the code for production readiness:

Identify the profiling tool or approach
Identify the bottlenecks using a profiling tool
Iteratively fix the code bottlenecks
Retest the functionality and performance

As deployer/Maintainer performance tune Reference Backend MongoDB Layer

Review and create Indexes for MongoDB Collections
Review and create references between existing collections in the Mongoose schema

Anganawadi Beneficiary Details Layout Design,ROI Generation and Verification

Anganawadi Benificiary Details Layout Design,ROI Generation and Verification

As a user i should be able to review the saved results for a student/layout

As a user, I should be able to review the saved results for a student/layout.

On the Save Status screen includes a Review button for each student/result
The review button should show up saved results stored in the backend for review.
Saved data for review gets fetched from the backend when
The user navigates to the student list page
The user saves data to the backend using Save All
The user should be able to close the review screen to go back to the Save Status Screen.
In Minimal Mode, the Save Data screen should have the same Review button to review save results.

Tentative Dev Completion Date: 5th Dec

ExamId to be populated by Saral App while invoking saveMarks API

As deployer/Maintainer I should be able to improve Handwritten Block letter Alphanumeric model Accuracy by experimenting with open and synthetic datasets - Iteration 2

As deployer/Maintainer I should be able to improve the Handwritten Block letter Alphanumeric model Accuracy by experimenting with open and synthetic datasets - Iteration 2.
Note: Expected accuracy improvement is approximately 3% to 5%.

Experiment with the existing model version (https://github.com/Sunbird-Saral/Project-Saral/tree/main/v1.0/frontend/SaralApp/android/saralsdk/src/main/assets#:~:text=model_letter_digits_finetune_18_07_22_epoch_15.tflite) to identify the miss-classifications.
Identifying open data sets
Experiment by inferencing the existing model with the new dataset and check the results and Prepare the new data
Training the next version of the model.
Inferencing and identifying optimal model version/checkpoint.
Package with SaralSDK/SaralReference App to test in real-time.

Selection of Set

Discussed in https://github.com/orgs/Sunbird-Saral/discussions/23

^{Originally posted by karsood September 22, 2022}
As of today, there are three selections that are made after logging in: Class, Section, and Subject.

With some states going for repeated examination using different sets (set A, set B, set C, set D), can we have sets come up as a selection after the subject is selected?

Benefit: This will help reduce selection error by the teacher.

If there are no sets for a round of examination in a state, the default entry can be A, as we have it for section selection.

Upgrade Async storage version in Saral App

As deployer/Maintainer I should be able to have all Saral ML Data preparation, Training, and Inferencing pipelines as part of the Saral repository

As deployer/Maintainer I should be able to have all Saral ML Data preparation, Training, and Inferencing pipelines as part of the Saral repository.

Creating folder structure in the repository for ml model assets and their pipelines
Place Training pipelines for Handwritten Digits, Alphanumeric Block English letter models.
Create Jupyter Notebooks for Handwritten Digits, Alphanumeric Block English letters experiments
Create Jupyter Notebooks for 2 versions of OCR detection.
Create overall process documentation for all the models.
Tentative Dev Completion Date: 5th Dec

This is to optimize the overall workflow and avoid intermediate /saveMarks calls during absent/present marking on the student list page.

Tentative Dev Completion Date: 28th Nov 2022

Saral offline mode support

To support Saral App in areas where internet coverage is limited, offline mode support is to be added:

Backend API call responses to be stored in local storage if the offline mode is enabled in /login API configuration.
After first call the data will be cached in local storage and avoids making subsequent API calls to the backend.
User to have provision to clear local storage and get the latest response from backend APIs
saveMarks,/login API needs internet connectivity to the backend APIs even in offline mode.

SaralSDK demo app scan results screen

App force update when a new version of app is available in playstore with backend config

App force updates when a new version of the app is available in the play store.

An option in the backend configuration with apkVersionId, and apkVersion details.
If Backend configuration is done to force an update, version details also need to be passed.

sunbird-saral / community Goto Github PK

community's People

Contributors

Stargazers

Watchers

Forkers

community's Issues

Discussed in https://github.com/orgs/Sunbird-Saral/discussions/12

Discussed in https://github.com/orgs/Sunbird-Saral/discussions/23

Recommend Projects

Recommend Topics

Recommend Org