Computer Vision in Python

OpenCV (installed as opencv-python) is the general-purpose base for computer vision: image and video I/O, filtering, geometric transforms, camera calibration and classic feature detection. Ultralytics packages YOLO models for detection, segmentation, pose estimation and classification behind a short train-and-predict API. Kornia reimplements many classical vision operations as differentiable PyTorch functions, which is useful for GPU augmentation inside training pipelines. FiftyOne helps you browse datasets and model predictions to find labeling mistakes and failure cases.

For OCR, pytesseract wraps the Tesseract engine, which you install separately; it works well on clean printed documents. EasyOCR uses deep learning models on PyTorch and is quick to set up for many languages and text in photos. PaddleOCR, built on PaddlePaddle, adds document parsing and layout analysis on top of multilingual recognition, which helps with forms, tables and scanned pages.