项目作者: general-labs

项目描述 :
General purpose AI experiments. Use face detection to identify politics figure, automatic face priority crop, image caption and labelling.
高级语言: Python
项目地址: git://github.com/general-labs/Image-AI.git
创建时间: 2019-09-22T18:06:19Z
项目社区:https://github.com/general-labs/Image-AI

开源协议:GNU General Public License v3.0

下载


Image AI

Image AI for content publishing platforms.

AI experiments for content publishing tools. Identify politics figures (obama, biden, trump, hillary etc). Automatically crop images by keeping subject at focus. Gather Image caption and labelling.

Goal

This is an experiment to utilize Image AI & Machine Learning in content publishing platforms such as CMS (Content management systems), curation tools, front end applications or even in backen API.

Features

  • Detect politicial figures (obama, biden, trump, hillary etc) using face detection.
  • Face priority image crop
  • Predict what image it is. Detect object and it’s name.
  • Predict what image it is using several different image datasets. Detect object and it’s name.
  • Detect scene or environment info.

=============================================

Deploy (Using Docker Compose)

cd docker

docker-compose up -d

Deploy (Native)

cd docker

pip install -r requirements.txt

cd ../scripts

python web-server.py

Visit http://localhost:5002.

Endpoints

Face priority image crop

curl 'http://localhost:5002/smart_crop?url=<YOUR_IMAGE_URL>'

Predict what image it is. Detect object and it’s name.

curl 'http://localhost:5002/predict_image?url=<YOUR_IMAGE_URL>'

Predict what image it is using several different image datasets. Detect object and it’s name.

curl 'http://localhost:5002/predict_image_all?url=<YOUR_IMAGE_URL>'

Detect scene or environment info.

curl 'http://localhost:5002/image_caption?url=<YOUR_IMAGE_URL>'

Detect scene or environment info.

curl 'http://localhost:5002/image_caption?url=<YOUR_IMAGE_URL>'

Detect politicial figures (person) using face detection.

curl 'http://localhost:5002/detect_person?url=<YOUR_IMAGE_URL>'

THIS IS WORK IN PROGRESS.

TO DO

  • …. huge list.

FULL CREDIT GOES TO EVERYONE INVOLVED IN ML/AI FIELD. WE WOULD BE IN STONE AGE WITHOUT THEIR DEDICATION AND HARD WORK.