01Cloud vision APIs
Google Cloud Vision
Google Cloud API for image labelling, object detection, text extraction and safe-search detection.
Best for: General-purpose image analysis on Google Cloud.
Visit Google Cloud Vision →Image Data Analytics Solutions
Vision APIs, computer vision platforms and open-source libraries that classify, detect, segment and describe the content of photographs, scans and satellite or medical images.
12Platforms & tools
3Categories
5FAQs answered
02Cloud vision APIs
Amazon Rekognition
AWS service that detects objects, scenes, text and faces in images and video.
Best for: Image analysis inside AWS.
Visit Amazon Rekognition →03Cloud vision APIs
Azure AI Vision
Microsoft's vision service for image analysis, captioning, text extraction and spatial analysis.
Best for: Image understanding in Azure applications.
Visit Azure AI Vision →04Cloud vision APIs
Imagga
Image recognition API for tagging, categorisation and visual search.
Best for: Auto-tagging image libraries.
Visit Imagga →05Computer vision platforms
Roboflow
Platform for annotating image datasets, training computer vision models and deploying them.
Best for: End-to-end custom vision projects.
Visit Roboflow →06Computer vision platforms
Landing AI
Computer vision platform known for data-centric AI and visual inspection use cases.
Best for: Industrial and manufacturing inspection.
Visit Landing AI →07Open-source libraries
OpenCV
The widely used open-source library for image processing and computer vision.
Best for: Classical image processing.
Visit OpenCV →08Open-source libraries
scikit-image
Python library of algorithms for image processing, built on NumPy and SciPy.
Best for: Scientific image analysis in Python.
Visit scikit-image →09Open-source libraries
Detectron2
Meta AI's library for object detection and segmentation research and production.
Best for: State-of-the-art detection and segmentation.
Visit Detectron2 →10Open-source libraries
MMDetection
Open-source object detection toolbox from the OpenMMLab project, built on PyTorch.
Best for: Comparing many detection models.
Visit MMDetection →11Open-source libraries
Ultralytics YOLO
Open-source real-time detection, segmentation and classification models with documentation.
Best for: Fast detection with simple tooling.
Visit Ultralytics YOLO →12Open-source libraries
Hugging Face Transformers
Library of pretrained models, including vision transformers, for images, text and audio.
Best for: Reusing pretrained vision models.
Visit Hugging Face Transformers →Choosing the right approach
Match your requirement to a starting point, then compare vendors.
| Tag or moderate uploaded photos | A cloud vision API such as Google Cloud Vision, Rekognition or Azure AI Vision |
|---|---|
| Train a custom detector | Roboflow, Landing AI or an open library such as Ultralytics or Detectron2 |
| Extract text from scans | OCR services or the multimodal OCR models directory |
| Research and prototyping | scikit-image, OpenCV or Hugging Face Transformers |
| Visual inspection in factories | A platform focused on inspection such as Landing AI |
| Large labelled datasets | Data labeling software for annotation and review |
Related pages on PeterIndia.net
Neighbouring directories and hubs.
Video Data Analytics SolutionsMoving-image analysis.Multimodal Language Models for OCRText from images and scans.Computer VisionThe computer vision hub.Data Labeling SoftwareAnnotate image datasets.Deep LearningNeural networks behind vision.Edge AIOn-device image analysis.Text Data Analytics SolutionsAnalyse text extracted from images.Data AnalyticsThe data analytics hub.
Frequently asked questions
Quick answers.
What is image data analytics?
It is extracting information from images: classifying what they show, detecting and segmenting objects, reading text, finding faces or defects, and analysing patterns across large image collections.
Do I need to train my own model?
Not always. Cloud APIs handle common tasks such as labelling and text extraction. Custom models help when your images are specialised, such as defects, medical scans or aerial imagery.
How much data is needed to train a detector?
It depends on the task and the model. Fine-tuning a pretrained model usually needs far fewer labelled images than training from scratch.
What about bias and privacy?
Image models can perform unevenly across groups and conditions, and images of people are personal data. Test on representative data and follow the rules that apply to you.
How does this relate to OCR?
OCR is a special case of image analysis that turns text in images into machine-readable text. See the multimodal OCR models directory.
Descriptions are summaries of each vendor's own positioning, checked against its website on 4 October 2026. Product names and offerings change often, so confirm details on the vendor site before deciding.