HCMAI2025_Baseline

A FastAPI-based AI application powered by Milvus for vector search, MongoDB for metadata storage, and MinIO for object storage.

🧑‍💻 Getting Started

Prerequisites

• Docker
• Docker Compose
• anaconda
• Python 3.10

Download the dataset

1. Embedding data and keys
2. Keyframes

Convert the global2imgpath.json to this following format(id2index.json)

{
  "0": "1/1/0",
  "1": "1/1/16",
  "2": "1/1/49",
  "3": "1/1/169",
  "4": "1/1/428",
  "5": "1/1/447",
  "6": "1/1/466",
  "7": "1/1/467",
}

🔧 Local Development

1. Clone the repo and start all services:

git clone https://github.com/Tuevu110405/Keyframe-retrieval.git
cd aio-aic

2. Install and setup env

conda create -n aic25
conda activate aic25
pip install -r requirements.txt

3. Run docker compose

docker compose up -d

4. Data Migration

python migration/embedding_migration.py --file_path <emnedding.pt file>
python migration/keyframe_migration.py --file_path <id2index.json file path>

5. Run the application

Open 2 tabs

5.1. Run the FastAPI application

cd gui
streamlit run main.py

5.1. Run the Streamlit application

cd app
python main.py