yyimg 1.0.2

image tools for deep learning

Introduction

yyimg is a high-level image-processing tool, written in Python and using OpenCV as backbend. This repo helps you with processing images for your deep learning projects.

Installation

Commands to install from pip or download the source code from our website https://pypi.org/project/yyimg

$ pip3 install yyimg==1.0.0rc

Example Useage

Take one image in Kitti dataset for example:

import yyimg
from PIL import Image
image, boxes, classes = yyimg.load_data()

Items	Description
image	a numpy array of shape (height, width, #channels)
boxes	a numpy array of shape (N, 5), representing N 2Dboxes of [class_index, xmin, ymin, xmax, ymax]
classes	a list of class names

print(classes)
['Car', 'Truck', 'Van', 'Pedestrian']

visualize 2D boxes

draw_image = yyimg.draw_2Dbox(image, boxes, class_category=classes)
draw_image = cv2.cvtColor(draw_image, cv2.COLOR_BGR2RGB) # BGR -> RGB
Image.fromarray(draw_image).show()

data augmentation

- horizontal_flip

with 2D bounding boxes:

aug_image, boxes = yyimg.horizontal_flip(image, boxes)

without 2D bounding boxes:

aug_image = yyimg.horizontal_flip(image)

- add_rain

aug_image = yyimg.add_rain(image)

- shift_gama

aug_image = yyimg.shift_gamma(image) 

- shift_brightness

aug_image = yyimg.shift_brightness(image)

- shift_color

aug_image = yyimg.shift_color(image)