Dogspotting

Use machine learning for computer vision to draw bounding boxes around dogs in pictures. See accompanying blog post.

NBViewer for easier viewing (recommended) here

Setup

1. (Optional) Create Anaconda environment w/ Python 3.6: $ conda create -n retinanet python=3.6 anaconda

2. (Optional) Activate virtual environment: $ source activate retinanet

3. Install dependencies: $ conda install tensorflow numpy scipy opencv pillow matplotlib h5py keras

4. Install ImageAI library: $ pip install https://github.com/OlafenwaMoses/ImageAI/releases/download/2.0.1/imageai-2.0.1-py3-none-any.whl

5. Download pretrained RetinaNet model: https://github.com/OlafenwaMoses/ImageAI/releases/download/1.0/resnet50_coco_best_v2.0.1.h5

Files

• Image_detection.ipynb: Jupyter notebook, walks through the steps and set up
• detect_single_image.py: Script to run to run object detection on a single image file. Set input_path to your image file.
• detect_images_folder.py: Script to run to run object detection on images in a folder. Set input_images_path to your-folder/*. Resulting images can be found in results/ directory. Keep in mind that the results folder images will be overwritten each time you run the script, so move previous files elsewhere.

Credits

• ImageAI library
• Object detection tutorial