docs: rewrite Computer Vision category intro

The intro predated the 2026-10-02 audit and did not mention the newly added paddleocr.

Co-Authored-By: Claude <noreply@anthropic.com>

website/data/category_intros/computer-vision.md

@@ -1,24 +1,27 @@
-Start with OpenCV when you need a Python computer vision library for images and video. To train and run detection models, use Ultralytics YOLO.
+OpenCV processes images and video. Training your own detector needs more than a Python computer vision library scoped to inference, so reach for Ultralytics.
 
 How to choose:
 
-- Image and video processing: OpenCV
-- Detection, segmentation, and pose models: Ultralytics YOLO
-- Vision ops inside a PyTorch model: Kornia
-- Dataset curation and model evaluation: FiftyOne
-- OCR on clean, printed documents: pytesseract
-- OCR on text in photos: EasyOCR
+- Reading, editing, and analyzing images and video: OpenCV
+- Detecting, segmenting, or tracking objects with a trained model: Ultralytics
+- Image operations inside a PyTorch model or training loop: Kornia
+- Browsing a dataset, finding label mistakes, and seeing where a model fails: FiftyOne
+- Printed text in clean scans: pytesseract
+- Text in photos or degraded scans: EasyOCR
+- Tables and page layouts, turned into Markdown or JSON: PaddleOCR
 
-OpenCV comes as [four pip packages that share the `cv2` namespace](https://github.com/opencv/opencv-python), so install only one: opencv-python for the main modules, or opencv-contrib-python to add the extra modules. If you never call `cv2.imshow` or you build your GUI with another toolkit, install the headless variant of either one, which also makes Docker images smaller.
+OpenCV's Python bindings [convert every OpenCV array to and from a NumPy array](https://github.com/opencv/opencv/blob/5.x/doc/py_tutorials/py_setup/py_intro/py_intro.markdown), so the images you load work with any other library that uses NumPy.
 
-Ultralytics YOLO covers the [whole life of a model](https://docs.ultralytics.com/modes/): train, validate, predict, export, and track, from Python or the `yolo` command. Its docs recommend [starting training from a pretrained model](https://docs.ultralytics.com/modes/train/). To deploy, [export it](https://docs.ultralytics.com/modes/export/) to ONNX, TensorRT, CoreML, or another format. The code and the models you train with it are [AGPL](https://www.ultralytics.com/license), so unless you open-source your whole project, you need an Enterprise License.
+Ultralytics runs YOLO models for detection, segmentation, pose estimation, and tracking from [one Python package and CLI](https://docs.ultralytics.com/). Load a pretrained model and [call `.train()` on your own dataset](https://docs.ultralytics.com/guides/finetuning-guide/): the weights carry over with no extra code. Ultralytics is AGPL, so [using it without open-sourcing your project](https://www.ultralytics.com/license) takes its Enterprise License.
 
-Kornia is a [differentiable computer vision library like OpenCV, with strong GPU support](https://kornia.readthedocs.io/en/latest/get-started/introduction.html). Every operator works on PyTorch tensors and supports autograd, so vision ops can run on the GPU and sit inside your training loop.
+Kornia is [a differentiable computer vision library like OpenCV, with strong GPU support](https://kornia.readthedocs.io/en/latest/get-started/introduction.html), built on PyTorch tensors. Its FAQ states the split: Kornia [provides operators to train networks, while OpenCV's scope is inference](https://kornia.readthedocs.io/en/latest/community/faqs.html).
 
-FiftyOne works on the data side of a model. [Load your dataset and your model's predictions into it](https://docs.voxel51.com/user_guide/basics.html), then see where the model succeeds and fails, and find mistakes in your labels. It [integrates with Ultralytics](https://docs.voxel51.com/integrations/ultralytics.html), so you can run and fine-tune YOLO models on FiftyOne datasets.
+FiftyOne is for building better datasets. [Load your images and labels as a Dataset](https://docs.voxel51.com/user_guide/basics.html) to search and sort samples and spot label mistakes. Then add your model's predictions and [evaluate them sample by sample](https://docs.voxel51.com/user_guide/evaluation/index.html) in the FiftyOne App to see where it goes wrong.
 
-pytesseract [wraps the Tesseract OCR engine](https://github.com/madmaze/pytesseract), which you install on its own, then put on your PATH or point `tesseract_cmd` at. Tesseract suits clean, printed text and needs no GPU. To get better results, [improve the image first](https://tesseract-ocr.github.io/tessdoc/ImproveQuality.html): Tesseract works best at 300 DPI or more, and retraining rarely helps unless you use an unusual font or a new language.
+pytesseract is [a wrapper for Google's Tesseract engine](https://github.com/madmaze/pytesseract), which you install on its own. `image_to_string()` returns the text, and `image_to_data()` adds boxes and confidences.
 
-EasyOCR is a [general OCR that reads text in photos as well as in documents](https://www.jaided.ai/easyocr), in dozens of languages. It handles text in photos, where Tesseract struggles, but it's slow without a GPU. Create a `Reader` for your languages [once](https://github.com/JaidedAI/EasyOCR) and reuse it for every image, since that call loads the model into memory.
+EasyOCR reads [many languages and scripts](https://github.com/JaidedAI/EasyOCR) with deep learning models on PyTorch. [Create one `Reader` and reuse it](https://www.jaided.ai/easyocr/tutorial/): it loads the models into memory once, then reads as many images as you give it.
 
-Ultralytics YOLO, Kornia, and EasyOCR run on PyTorch. When you need a specific CUDA build, install PyTorch before the library, as [Ultralytics](https://docs.ultralytics.com/quickstart/) and [Kornia](https://kornia.readthedocs.io/en/latest/get-started/installation.html) recommend. EasyOCR's README says the same for Windows.
+PaddleOCR does two jobs: [general OCR](https://www.paddleocr.ai/latest/en/version3.x/pipeline_usage/OCR.html) reads text, and document parsing [turns PDFs and document images into Markdown and JSON](https://www.paddleocr.ai/latest/en/index.html) that keep the original structure.
+
+Load and crop images with OpenCV, then pass the NumPy arrays on: [EasyOCR](https://www.jaided.ai/easyocr/tutorial/), [PaddleOCR](https://www.paddleocr.ai/latest/en/version3.x/pipeline_usage/OCR.html), and [Ultralytics](https://docs.ultralytics.com/modes/predict/) all take them as input.