/cvlib

A high level, easy to use, open source Computer Vision library for Python.

Primary LanguagePythonMIT LicenseMIT

Downloads Gitter

cvlib

A high level easy-to-use open source Computer Vision library for Python.

Installation

Provided the below python packages are installed, cvlib is completely pip installable.

  • numpy
  • opencv-python
  • requests
  • progressbar
  • pillow
  • tensorflow
  • keras

Install the required packages using the below command

pip install -r requirements.txt

pip install cvlib

To upgrade to the newest version pip install --upgrade cvlib

If you are using a GPU, edit the requirements.txt file to install tensorflow-gpu instead of tensorflow.

Note: Python 2.x is not supported

Face detection

Detecting faces in an image is as simple as just calling the function detect_face(). It will return the bounding box corners and corresponding confidence for all the faces detected.

Example :

import cvlib as cv
faces, confidences = cv.detect_face(image) 

Seriously, that's all it takes to do face detection with cvlib. Underneath it is using OpenCV's dnn module with a pre-trained caffemodel to detect faces.

Checkout face_detection.py in examples directory for the complete code.

Sample output :

Gender detection

Once face is detected, it can be passed on to detect_gender() function to recognize gender. It will return the labels (man, woman) and associated probabilities.

Example

label, confidence = cv.detect_gender(face)

Underneath cvlib is using a pre-trained keras model to detect gender from face. The accuracy is not so great at this point. It still makes mistakes. Working on adding a more accurate model.

Checkout gender_detection.py in examples directory for the complete code.

Sample output :

Object detection

Detecting common objects in the scene is enabled through a single function call detect_common_objects(). It will return the bounding box co-ordinates, corrensponding labels and confidence scores for the detected objects in the image.

Example :

import cvlib as cv
from cvlib.object_detection import draw_bbox

bbox, label, conf = cv.detect_common_objects(img)

output_image = draw_bbox(img, bbox, label, conf)

Underneath it uses YOLOv3 model trained on COCO dataset capable of detecting 80 common objects in context.

Checkout object_detection.py in examples directory for the complete code.

Sample output :

License

cvlib is released under MIT license.