AceTrackAI - A Student Feedback Generator

AceTrackAI is an AI-powered tool designed to evaluate student responses to historical questions and provide constructive, context-aware feedback. Developed using T5 transformers, the project automates the assessment process, offering actionable insights to students and instructors.

---

Features

• Upload PDF Exams: Upload scanned exam papers directly for evaluation.
• Automated Parsing: Extracts questions, responses, and reference answers from the uploaded documents.
• AI-Powered Feedback: Generates accurate feedback based on predefined correct answers using a fine-tuned T5 model.
• User-Friendly Interface: Built with Streamlit, ensuring simplicity and accessibility.

---

Installation

1. Clone the repository:

   git clone https://github.com/Abmstpha/AceTrackAI.git
   cd AceTrackAI

2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # For Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Ensure the sentencepiece library is installed for T5 tokenizer compatibility:

   pip install sentencepiece

Usage

1. Prepare the Model Weights: The repository does not include .pt files due to size constraints. Please download the model weights separately and place them in the appropriate directory.

2. Start the backend:

   python app.py

3. Start the frontend:

   streamlit run frontend.py

4. Access the application: Open your browser and navigate to http://localhost:8501.

5. Upload a PDF of the exam and view the AI-generated feedback.

Project Structure

AceTrack/
├── app.py                 # Backend script
├── frontend.py           # Frontend script
├── requirements.txt      # List of dependencies 
├── model weights (PT file not included due to size) 
├── AceTrack_T5_weights/  # Directory for 
│   ├── special_tokens_map.json
│   ├── tokenizer_config.json
│   ├── spiece.model
│   └── added_tokens.json
└── README.md             # This README file