Update README.md

README.md

@@ -1,39 +1,73 @@
-# INLCUDE - Isolated Indian Sign Language Recognition
+# INCLUDE - Isolated Indian Sign Language Recognition with Web UI
 
-This repository contains code for training models on [INCLUDE](https://zenodo.org/record/4010759) dataset
+This repository features an enhanced implementation of the **INCLUDE** dataset for Indian Sign Language (ISL) recognition. It introduces modern deep learning architectures, a comprehensive preprocessing pipeline, and a full-stack web interface for real-time video inference.
 
-# Dependencies
+---
 
-Install the dependencies through the following command
+## 🚀 Getting Started
+
+### 1. Environment Setup
+Install the necessary Python dependencies:
 
 ```bash
->> pip install -r requirements.txt
+pip install -r requirements.txt
 ```
 
+### 2. Web Interface Setup
+The UI is built with React. To prepare the frontend:
 
+```bash
+cd ui
+npm install
+npm run build  # Builds the UI for production
+```
+
+For local development:
 
-## Steps
-- Download the INCLUDE dataset
-- Run `generate_keypoints.py` to save keypoints from Mediapipe Hands and Blazepose for train, validation and test videos. 
 ```bash
->> python generate_keypoints.py --include_dir <path to downloaded dataset> --save_dir <path to save dir> --dataset <include/include50>
+npm run dev
 ```
-- Run `runner.py` to train a machine learning model on the dataset
+
+### 3. Running the Application
+Launch the FastAPI server to start the web interface:
+
 ```bash
->> python runner.py --dataset <include/include50> --use_augs --model transformer --data_dir <location to saved keypoints>
+python app.py
 ```
-- Use the `--use_pretrained` flag to either perform only inference using pretrained model or resume training with the pretrained model. 
+
+Access the interface at http://localhost:8000.
+
+---
+
+## 📈 Training and Evaluation
+
+### Extract Keypoints
+Process the INCLUDE dataset to extract pose and hand keypoints:
+
 ```bash
->> python runner.py --dataset <include/include50> --use_augs --model transformer --data_dir <location to saved keypoints> --use_pretrained <evaluate/resume_training>
+python generate_keypoints.py --include_dir <path_to_dataset> --save_dir <path_to_save> --dataset include50
 ```
-- To get predictions for videos from a pretrained model, run the following command.
+
+### Train a Model
+To train a Transformer-based model:
+
 ```bash
->> python evaluate.py --data_dir <dir with videos>
+python runner.py --dataset include50 --use_augs --model transformer --data_dir <keypoints_dir>
 ```
 
-## Citation
+### Perform Inference
+Run predictions on a folder of videos:
 
+```bash
+python evaluate.py --data_dir <path_to_videos>
 ```
+
+---
+
+## 📝 Citation & Credits
+This project is built upon the INCLUDE dataset and research by the following authors:
+
+```bibtex
 @inproceedings{10.1145/3394171.3413528,
 author = {Sridhar, Advaith and Ganesan, Rohith Gandhi and Kumar, Pratyush and Khapra, Mitesh},
 title = {INCLUDE: A Large Scale Dataset for Indian Sign Language Recognition},
@@ -41,13 +75,8 @@ year = {2020},
 isbn = {9781450379885},
 publisher = {Association for Computing Machinery},
 doi = {10.1145/3394171.3413528},
-numpages = {10},
 series = {MM '20}
 }
 ```
 
-# INCLUDE
-# INCLUDE
-# INCLUDE
-# INCLUDE
-# INCLUDE
+> **Note:** This repository includes custom UI development and model architectural enhancements beyond the original publication.