CosmeticDetectionYolo3

Introduction

A Cosmetic detector using YOLOv3 (Tensorflow backend) for cosmetic detection (ref: qqwweee/keras-yolo3)

Requirement

Environment

Python 3.6.8
Keras 2.2.4
Tensorflow 1.6.0
CUDA 9.0
cudnn 7.0
opencv 3.4.0

Libraries

colormath
Pillow
numpy
Keras
opencv-python
imgaug
matplotlib

Test data

Test data is under 'images/'

How to use:

1) Get the model

Step 1: Download the project:

git clone https://github.com/2013-11390/cosmeticDetectionYolo3.git

Step 2: Download YOLOv3 weights from YOLO website or yolov3.weights.

Step 3: Convert the Darknet YOLO model to a Keras model

python convert.py yolov3.cfg yolov3.weights model_data/yolo.h5	# to get yolo.h5(model)

OR download the model yolo.h5 to model_data/ directory directly.

2) Test the model on coco dataset(original yolo model is trained on coco dataset)

Run YOLO detecion.

python yolo_video.py --model_path model_data/yolo.h5 --classes_path model_data/coco_classes.txt --image

3) Train the model for cosmetic detection:

Step 1: Put cosmetic image in images folder by name 'image.jpg'

Step 2: Augment test data

python imgAug.py

Step 3: Train the model(use yolo.h5 as the pretrained model)

python train.py -a imagesAug/train.txt -c model_data/coco_classes.txt -o model_data/custom_cosmetic_coco.h5

Step 4: Run the model

python yolo_video.py --model_path model_data/custom_cosmetic_coco.h5 --classes_path model_data/coco_classes.txt --input 'your video name'

if you want to save the result of yolo object detecting

python yolo_video.py --model_path model_data/custom_cosmetic_coco.h5 --classes_path model_data/coco_classes.txt --input 'your video name' --output 'save file name'

HONUBIN/CosmeticDetectionYolo3-version1-