Contact person: Business Client

Phone:Show

Email:Show

Location: Askari 2, Pakistan

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"Freelancer.com Job Posting — Full Project Brief for HandLingo (Intelligent Sign Language Agent)
Budget: PKR 100,000
Technologies: Android (Java/Kotlin), TensorFlow/TFLite, Python (ML), FastAPI/Node.js, SQL/NoSQL, Cloud Deployment, Figma, Web Dashboard
1. Project Overview

HandLingo is a mobile application designed to translate Pakistani Sign Language (PSL) in two directions:

1. Text → Sign (Dictionary Mode)

User enters a word (Urdu/English).

App plays the corresponding sign video/animation.

Videos must be cached for offline use.

2. Sign → Text (Recognition Mode)

User shows a PSL gesture to the camera.

Model predicts the label, shows confidence scores, allows correction, and saves history.

This project includes:

Android mobile app

Machine learning model (PSL recognition)

Backend API + database

Admin dashboard

Dataset preparation

Deployment

Testing + documentation

The complete SRS document (IEEE-based) is provided and must be followed exactly, though the freelancer may propose tooling/method improvements.

2. Full Scope of Work (Everything to Be Outsourced)
A. Android Mobile App Development

Build the full Android app (Kotlin/Java).

Implement both modes:

Dictionary (Text → Sign with video playback)

Recognition (Sign → Text using ML model)

Multilingual UI:

English

Urdu (RTL support)

Key screens:

Splash + Onboarding

Home (Mode Selection)

Dictionary (search, favorites, video)

Camera + Recognition screen

Results + corrections

History

Settings

TTS for English + Urdu.

Offline caching of dictionary videos.

Error handling (low light, permission issues, invalid inputs).

Integration with backend APIs and TFLite model.

B. Machine Learning Model Development

Build a full PSL recognition model for static signs.

Dataset creation workflow:

Obtain PSL dataset (from [login to view URL] or any verified source).

Clean, label, and preprocess data.

Minimum: 28+ classes; ideally expandable to 50–100 signs.

Preprocessing:

Image normalization

Resizing

Augmentation

Model architecture:

Baseline CNN or MobileNetV2 (freelancer’s expertise allowed)

Must achieve ≥85% validation accuracy

Convert to TFLite + quantization.

Provide model versioning + metadata.

C. Backend API Development

Backend should include:

1. Authentication

JWT-based login/register

Phone/email support

2. Dictionary APIs

Search

Fetch video

Upload new sign videos (admin)

3. Recognition API

Optional: support server-side inference for low-powered devices.

4. Feedback Pipeline

Save corrected labels

Store user-submitted misclassification images

5. Analytics

Usage metrics

Popular signs

Failure cases

D. Admin Dashboard (Web)

Upload PSL sign videos

Manage dictionary database

View analytics (charts, summary)

Model version page

Feedback review page

User management (roles: user/admin)

Tech stack flexible (React, [login to view URL], Vue, etc.)

E. Database & Cloud Infrastructure

Database schema as per SRS (Users, Dictionary, Models, Feedback, etc.)

Secure API hosting

HTTPS/TLS, JWT, password hashing

Storage for sign videos + model artifacts

Deployment (AWS/DigitalOcean/Render/Heroku)

F. Testing & QA

Functional testing

Performance testing

Security testing

Mobile device testing (mid-range Android phones)

UAT scripts

G. Documentation

Complete source code documentation

API documentation (Swagger/Postman)

ML model documentation

Deployment guide

App user manual

Admin panel manual

3. Deliverables (Everything You Must Provide)

Freelancer must deliver all of the following:

A. Mobile App Deliverables

APK (debug + production builds)

Complete Android Studio project

UI/UX assets (icons, animations, Figma designs)

Integrated TFLite model

Offline dictionary support

Full multilingual support (Urdu/English)

B. Machine Learning Deliverables

PSL dataset (raw + cleaned)

Preprocessing scripts (Python)

Full Jupyter notebooks for training

Trained model (.h5 or .pt)

TFLite model (.tflite)

Model metadata (accuracy report, confusion matrix)

Model version control setup

C. Backend Deliverables

Fully functional REST API

Complete source code

Database schema + migrations

Authentication module

Dictionary management API

Recognition API

Feedback collection system

Logging + monitoring setup

Deployed backend URL

Admin dashboard URL

D. Admin Dashboard Deliverables

Admin login

Upload sign videos

Manage dictionary

View analytics dashboard

View model stats

Review feedback samples

Source code (React/[login to view URL])

E. Deployment Deliverables

Backend deployed

Frontend/admin deployed

Database configured

Video storage configured

SSL-enabled domain

CI/CD (optional but preferred)

F. Documentation Deliverables

Technical documentation

User manual (PDF)

Admin dashboard guide

API documentation (Swagger/Postman Collection)

Testing report

Final project report (optional but preferred)

4. Responsibilities of the Freelancer

Develop end-to-end full system as per SRS.

Provide regular updates (minimum weekly).

Deliver complete source code with no hidden dependencies.

Maintain confidentiality (NDA if required).

Provide 1 month of post-delivery support.

5. Skills Required

Android (Kotlin/Java)

Python (TensorFlow / PyTorch)

TensorFlow Lite (TFLite optimization)

Computer Vision & CNN expertise

Backend (FastAPI / Node.js / Django / Express)

Database (MySQL / PostgreSQL / MongoDB)

DevOps & deployment (AWS/Render/DigitalOcean)

React / [login to view URL] for admin dashboard

UX/UI design (Figma)

6. Acceptance Criteria

App must run smoothly on mid-range Android devices.

Recognition accuracy ≥ 85%

Video playback must start under 1 second (cached).

Recognition latency ≤ 1.5 seconds

Admin uploads appear in dictionary within 5 minutes.

Urdu + English UI supported.

SRS must be fully implemented.

7. Freelancer Prompt (To Attach Inside Project Files)

“You are hired to build HandLingo — a complete PSL recognition and dictionary mobile application. Follow the approved SRS document provided. You may propose changes in tools, libraries, and implementation approaches, but you MUST NOT remove or reduce any deliverable or mandatory requirement from the SRS.
You are responsible for the Android app, backend, admin dashboard, database, ML model, dataset creation, deployment, documentation, and testing.
You must deliver the entire system fully functional, stable, and production-ready within the agreed timeline and budget. All source code and assets will be the client’s intellectual property.”" (client-provided description)