Computer Vision Libraries
OpenCV, Pillow, scikit-image, PyTorch vision tools and annotation platforms.
This is an overview — pick a sub-topic below.
Sub-topics (16)
OpenCV (cv2)
The most widely used computer-vision library: image processing, camera capture, features, calibration and DNN inference.
Pillow
The friendly Python Imaging Library for opening, editing and saving images.
scikit-image and Mahotas
Scientific image-processing algorithms on NumPy arrays (filters, segmentation, measurement).
SimpleITK and ITK
Medical image processing and registration toolkits.
Kornia
Differentiable computer-vision operations in PyTorch (augmentation, geometry, filters) that run on GPU.
Albumentations and imgaug
Fast image augmentation for training (flips, crops, colour changes, blur) that also transforms boxes and masks.
torchvision and timm
Datasets, transforms and pretrained models for PyTorch (torchvision); hundreds of state-of-the-art backbones (timm).
Detectron2 and OpenMMLab
Research-grade detection and segmentation frameworks from Meta (Detectron2) and OpenMMLab (MMDetection, MMSegmentation, MMPose, MMTracking, MMPreTrain, MMEngine, MMOCR).
FiftyOne, CVAT, Supervisely and Roboflow
Dataset curation and annotation tools: label images, inspect predictions, find bad labels.
supervision, SAHI and Norfair
Utilities around detectors: drawing/counting zones (supervision), slicing for small objects (SAHI), lightweight tracking (Norfair).
Open3D, Trimesh, PyVista, PyMeshLab, pyrender, vedo and PyTorch3D
3-D data processing: point clouds, meshes, rendering and differentiable 3-D.
MediaPipe
Google's ready-made real-time pipelines for hands, face mesh, pose and objects.
DeepFace, InsightFace and face_recognition
Face detection, verification and attribute analysis libraries.
Deep SORT, ByteTrack, BoT-SORT and motpy
Multi-object tracking algorithms that assign stable IDs to detections across frames.
PaddleOCR, EasyOCR, Tesseract, docTR, keras-ocr, MMOCR and OCRmyPDF
OCR engines and tools — from classic Tesseract (pytesseract) to deep-learning OCR, plus adding a text layer to scanned PDFs (OCRmyPDF).
DeepStream Python Bindings
NVIDIA's video-analytics SDK for multi-camera pipelines on Jetson and GPUs.
Spotted a mistake or want to add detail? Suggest an improvement — Content Contributors' edits are reviewed by editors before they go live.
Discussion (0)
Sign in or join free to ask questions and discuss this topic.