Skip to content

Latest commit

 

History

History
72 lines (53 loc) · 2.53 KB

README.md

File metadata and controls

72 lines (53 loc) · 2.53 KB
Raw

ezimage

What is ezimage?

Load and display images and access its content data with one-liners. A wrapper for PIL and IPython.display, ideal for machine learning and image processing. Easily load a PNG/JPEG image (or a list of them) from a local machine or the web and use its data in any desired format (e.g. NumPy/PyTorch/TensorFlow/etc). Display or save data arrays from any format or ordering (CHW/WCH). All with one-liners.

Installation

pip install ezimage

Example Usage

from ezimage import ezimage
img = ezimage('https://tinyurl.com/yyxvexs6')
img.CHW
img.display()

Watch this tutorial on YouTube for a walk through.

Description

Enables the user to easily load an image from a path, a url, or by directly providing a data array. If a list is provided, a list of ezimage instances will be retunred. The properties enable the user to access the image data in R, G, B, A, GS (grayscale), HWC, and CHW formats. In IPython environments such as in Jupyter, the display method allows the user to view images. Unlike PIL, this method allows for iterative image display.

Regardless of the ordering, format, or type of data, if directly feeding data arrays, all values must be in a valid PNG range (between 0 and 255).

Versatile Data Type and Data Format

User is free to choose the format of the data arrays specified with format_func.

format_func is a function that converts a PIL image to user's desired format. The output must cast to a NumPy array (i.e. no error when running np.array(format_func(PIL_img))). See examples below.

Note that the internal computations are done using PIL functions or NumPy arrays and format_func is only designated for the user interface.

NumPy uint8 Example:

format_func=lambda PIL_Image: np.array(PIL_Image, dtype=np.uint8)

NumPy float16 Example:

format_func=lambda PIL_Image: np.array(PIL_Image, dtype=np.float16)

PyTorch float32 Example:

import torch
format_func=lambda PIL_Image: torch.tensor(np.array(PIL_Image), dtype=torch.float32)

Example usage for a single image from the web:

img = ezimage("https://upload.wikimedia.org/wikipedia/commons/thumb/d/da/Omar_Khayyam2.JPG/220px-Omar_Khayyam2.JPG")
img.display()

Example usage for reading all PNG files in a directory:

from ezimage import ezimage
from pathlib import Path
path_parent = Path("path/to/image/folder/")
pathList = list(path_parent.glob("*.png"))
imgList = [ezimage(p) for p in pathList]
for img in imgList:
    img.display(print_name=True)