/EyeQ-Foundation

Primary LanguagePythonMIT LicenseMIT

EyeQ-Foundation

EyeQ-Foundation is a repository contains some zero shot computer vision models. The aim of this project is to make process automatic. For example automatic label service.

Funcationalities:

  • Zero Shot Object Detection
  • Auto Label image from directory
  • REST API service
  • Pip installation
  • Docker installatin for REST Server using FastAPI
  • CLIP to correctly identify the labels

Available Models:

  • Segment Anything
  • Grounding Dino

TODO:

  • Segment Anything HQ
  • Batched Inference
  • RAM
  • Tag2Text
  • Knowledge Distillation application
  • CLIP

How to use

Pip Package:

Installation:

pip install git+https://github.com/facebookresearch/segment-anything.git
pip install git+https://github.com/IDEA-Research/GroundingDINO.git
pip install git+https://github.com/openai/CLIP.git

Usage in code:

from eyeq_foundation import SAM, Dino
teacher_model = SAM()

Docker Service:

Build :

docker build -t eyeq-foundation -f docker/Dockerfile-cpu .

Start:

docker run -d --name eyeq-foundation -p 9000:9000 eyeq-foundation

Logs:

docker container logs eyeq-foundation