https://github.com/anuradhawick/ai-album
A template project that puts together BLIP image captioning and facial feature extraction to emulate a simple web based photo manager.
https://github.com/anuradhawick/ai-album
facial-recognition image-captioning
Last synced: about 2 months ago
JSON representation
A template project that puts together BLIP image captioning and facial feature extraction to emulate a simple web based photo manager.
- Host: GitHub
- URL: https://github.com/anuradhawick/ai-album
- Owner: anuradhawick
- License: gpl-3.0
- Created: 2022-10-22T15:18:56.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2023-04-25T06:17:30.000Z (about 2 years ago)
- Last Synced: 2025-02-05T11:18:32.930Z (4 months ago)
- Topics: facial-recognition, image-captioning
- Language: Python
- Homepage:
- Size: 8.38 MB
- Stars: 0
- Watchers: 3
- Forks: 3
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# AI-Album
Hobby project in progress. Everything is still in progress. Re-use at your own risk!
### Implemented features
- Image indexing (recursive album creation and loading to database)
- Image captioning with multiple GPUs
### Upcoming features
- Image face annotation
- Face index
# Dependencies
## python dependencies
```bash
mamba create -n aialbum transformers python=3.10 pytorch fairscale dask-mongo torchaudio pytorch-cuda=11.7 iopath cudatoolkit=11.7 -c pytorch -c nvidia -c iopath -c conda-forge
mamba install rapids=23.02 -c rapidsai -c conda-forge -c nvidia
pip install "fastapi[all]" pillow pillow-heif einops spacy pycocoevalcap cryptography==38.0.4 motor pymongo pyyaml networkx omegaconf timm decord opencv-python webdataset jupyterlab torchvision
pip install tensorflow
pip install gdown
```
Installation in Raspberry Pi 4 B (8 GB)
```bash
mamba create -n aialbum transformers python=3.10 pytorch dask-mongo iopath -c pytorch -c iopath -c conda-forge
pip install fairscale torchaudio "fastapi[all]" pillow pillow-heif einops spacy pycocoevalcap cryptography==38.0.4 motor pymongo pyyaml networkx omegaconf timm opencv-python webdataset jupyterlab torchvision
```
## post install
```bash
python -m spacy download en_core_web_sm
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CONDA_PREFIX/lib/
```
## Building DLIB library
```bash
./build.sh
```
# Development
## Docker for mongodb
```bash
export UID=$(id -u)
export GID=$(id -g)
docker-compose up -d mongo
```
## Running the dev server
```
uvicorn server.__main__:create_app --factory --reload
```
# Deployment
```
python -m server
```
# Clean DB entries
```bash
python server/clean.py
```