To see more details refer to this official documentation Detail readme

The repository includes:

• To setup this repo clone it first

• After cloning is done create a new environment using

  conda create -n mask_rcnn python=3.6

• The activate the environment using

  conda activate mask_rcnn

• Once envrinment is activated then use

  pip install -r requirements.txt

• At this point you should have a working environment

• Download mask_rcnn_balloon.h5. Save it in the root directory of the repo (the mask_rcnn directory) from here

• demo.ipynb can be used to see the output of the pretrained model on the images

• All the images for testing are under aetrex folder

• Inside Run Object Detection section set the image directory path, then it choose the image from there randomly and show the predictions like below

  IMAGE_DIR='../aetrex/images'

• Note - skimage.io.imread() gives output image in 4 channels sometime as oppose to opencv which give 3 channel image by default, if you get channel error than use skimage.io.imread()[:,:,:3]