Skip to main content
Ryan Orban

Ryan Orban

Subject
7 entries

Vision

Bookmarks

  1. A Visual Guide to Vision Transformers

    A visual guide to Vision Transformers (ViT) — explains how the transformer architecture is adapted for images, covering patch embeddings, position encodings, and attention in visual domains with diagrams. Good complement to the original ViT paper for building intuition.

  2. macOSpilot: Voice + Vision AI Assistant for macOS

    macOSpilot is an open-source macOS AI assistant that combines voice input and screen vision to answer questions about any application in context — it sees what you're looking at and hears your question, then responds in audio. An early demonstration of multimodal desktop AI assistance.

  3. Bulktag: Batch Image Tagging with GPT-4 Vision

    Bulktag uses GPT-4 Vision to batch-tag images automatically — send a folder of images, get back descriptive tags for each. An early practical application of GPT-4V for metadata generation and image library organization.

  4. EVA: Exploring the Limits of Masked Visual Representation Learning at Scale

    EVA is a 1-billion-parameter vision foundation model from BAAI that achieves state-of-the-art on image classification, detection, and segmentation by pretraining a ViT to reconstruct masked CLIP features. Initializing CLIP's vision tower from EVA dramatically stabilizes training — an important practical finding for building large multimodal systems.

  5. Generative Pretraining from Pixels (iGPT)

    The iGPT paper from OpenAI (ICML 2020) showing that a GPT-2-scale transformer trained to autoregressively predict pixels learns strong image representations — 96.3% accuracy on CIFAR-10 with a linear probe. It's a direct transposition of NLP pretraining ideas to the image domain, predating CLIP and DALL-E.

  6. Eigengrau

    Eigengrau is the dark gray color perceived by the human visual system in the absence of light — not true black, but a specific gray around #16161d. The visual cortex generates this background noise spontaneously, which is why total darkness doesn't look fully black.

  7. Biologists Grow Human-Eye Precursor from Stem Cells

    Nature News coverage of a 2012 milestone: biologists grew a human eye cup (optic cup) from induced pluripotent stem cells, self-organizing into the correct layered structure. An early demonstration that complex organ structures could self-assemble from stem cells — significant for regenerative medicine and our understanding of developmental biology.

All bookmarks