Common Python Libraries for Computer Vision Projects

In this article we present a curated list of widely used Python libraries for computer vision, offering brief overviews of their capabilities, typical use cases, and example code snippets to help developers get started quickly.

1. Pillow

Pillow is a user‑friendly Python imaging library that supports opening, manipulating, and saving many image formats, offering basic operations such as cropping, resizing, rotating, color changes, and adding text or shapes. It is also the image processing backend used by torchvision.

2. OpenCV (Open Source Computer Vision Library)

OpenCV is one of the most popular image‑processing libraries, originally developed by Intel and now widely used in computer vision. It provides a large collection of algorithms for vision and machine learning, is highly optimized for real‑time applications, and reads images in BGR format.

cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

3. Mahotas

Mahotas offers a set of high‑performance image processing functions implemented in C++ with multithreading. It includes morphological operations, segmentation, and distance transforms, providing a simpler API than OpenCV while maintaining comparable speed.

Example using watershed segmentation:

# import using ``mh`` abbreviation which is common:
import mahotas as mh

# Load one of the demo images
im = mh.demos.load('nuclear')

# Automatically compute a threshold
T_otsu = mh.thresholding.otsu(im)

# Label the thresholded image
seeds, nr_regions = mh.label(im > T_otsu)

# Call seeded watershed to expand the threshold
labeled = mh.cwatershed(im.max() - im, seeds)

Simple distance map example:

import pylab as p
import numpy as np
import mahotas as mh

f = np.ones((256, 256), bool)
f[200:,240:] = False
f[128:144,32:48] = False

dmap = mh.distance(f)
p.imshow(dmap)
p.show()

4. Scikit‑Image

Built on top of scikit‑learn, scikit‑image provides a comprehensive set of algorithms for image segmentation, geometric transformations, color space operations, and filtering, with support for multi‑dimensional images and seamless integration with NumPy and SciPy.

from skimage import data, io, filters
image = data.coins()
edges = filters.sobel(image)
io.imshow(edges)
io.show()

5. TensorFlow Image

TensorFlow Image is a TensorFlow module that handles image decoding, encoding, cropping, resizing, and conversion, leveraging GPU acceleration for large datasets. It can be used as part of a training pipeline via Keras utilities.

batch_size = 32
img_height = 180
img_width = 180

train_ds = tf.keras.utils.image_dataset_from_directory(
    data_dir,
    validation_split=0.2,
    subset="training",
    seed=123,
    image_size=(img_height, img_width),
    batch_size=batch_size)

6. PyTorch Vision

PyTorch Vision is the image‑processing component of the PyTorch ecosystem, offering datasets, models, and utilities for computer‑vision tasks.

import torchvision
video_path = "path to a test video"
reader = torchvision.io.VideoReader(video_path, "video")
reader_md = reader.get_metadata()
print(reader_md["video"]["fps"])
video.set_current_stream("video:0")

7. SimpleCV

SimpleCV builds on OpenCV, PIL, and NumPy to provide a high‑level API for loading, processing, and analyzing images, aimed at beginners and non‑experts.

import SimpleCV
camera = SimpleCV.Camera()
image = camera.getImage()
image.show()

8. Imageio

Imageio offers a simple API for reading and writing a wide range of image and video formats, supporting NumPy arrays, PIL images, and byte strings, and includes frame‑by‑frame video handling.

import imageio.v3 as iio
im = iio.imread('imageio:chelsea.png')  # read a standard image
im.shape  # (300, 451, 3)
iio.imwrite('chelsea.jpg', im)  # convert to jpg

9. Albumentations

Albumentations is a fast, flexible library for image and data augmentation, widely used in machine‑learning pipelines and capable of handling masks and bounding boxes.

import albumentations as A
import cv2

transform = A.Compose([
    A.RandomCrop(width=256, height=256),
    A.HorizontalFlip(p=0.5),
    A.RandomBrightnessContrast(p=0.2),
])

image = cv2.imread("image.jpg")
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
transformed = transform(image=image)
transformed_image = transformed["image"]

10. timm

timm is a PyTorch model library that provides a large collection of pretrained computer‑vision models, useful for deep‑learning experiments.

import timm
import torch
model = timm.create_model('resnet34')
x = torch.randn(1, 3, 224, 224)
model(x).shape

Regardless of whether you are starting with basic image processing or exploring advanced machine‑learning models, these libraries offer essential tools for a wide range of computer‑vision tasks.