iOS + YOLOv3

Introducing online object detection on iOS with YOLOv3-416 and YOLOv3-tiny convolutional neural network architecture.

Convertation from Darknet to CoreML

First of all need to download YOLOv3 pretrained weights from YOLO website. Download both cfg and weights files.

Then load Darknet weights to Keras model using Keras-YOLOv3 implementation.

After cloning above repo use this commend to load Darknet and save .h5:

python convert.py yolov3.cfg yolov3.weights model_data/yolo.h5

And finally to transform from .h5 keras model representation to CoreML format use code below:

import coremltools

coreml_model = coremltools.converters.keras.convert(
    'yolo.h5',
    input_names='image',
    image_input_names='image',
    input_name_shape_dict={'image': [None, 416, 416, 3]},
    image_scale=1/255.)

coreml_model.license = 'Public Domain'
coreml_model.input_description['image'] = 'Input image'

coreml_model.save('yolo.mlmodel')