A phonebook for pre-trained computer vision models

What it does

This repository is a curated directory of pre-trained computer vision models, sorted into markdown tables by framework. It links to external implementations for tasks like object detection, segmentation, super-resolution, and image captioning, and tags each entry with its license. It is essentially an index that saves you from searching each framework’s ecosystem separately.

The interesting bit

The README is almost entirely markdown tables linking elsewhere; the value is in the aggregation, not in original code. The maintainer has corralled everything from mainstream Keras applications to niche experiments like brain-tumor U-Net into one skim-friendly index.

Key highlights

• Covers five major frameworks: TensorFlow, Keras, PyTorch, Caffe, and MXNet.
• Includes license information for most entries, though several are marked “Not Found.”
• Links to sibling lists for NLP and audio pre-trained models.
• Recommends Netron for visualizing network architectures.
• Spans classic classifiers (VGG16, ResNet), detectors (YOLO, Mask R-CNN), and generative models (pix2pix, CycleGAN).

Caveats

• A number of entries carry a “Not Found” license warning, leaving legal clearance as an exercise for the reader.
• The introductory text contains typos (“some one else”, “self learning car”), suggesting the curation is a volunteer effort rather than an institutional catalog.
• This is purely a link directory; expect no installable package, unified API, or code beyond the README tables.

Verdict

Useful if you need a quick survey of available pre-trained models across frameworks. Not for you if you want a unified, downloadable model zoo or original training code.