visual-memory

ReflectWorld-MM: An Entity-Oriented Multimodal Memory System for Open-Ended Video Stream

Native-video memory for vision-language-action models, using timestamped visual history and exact streaming inference for long-horizon robot manipulation.

Official implementation of "Beyond the Timeline: Augmenting Long-Video Memory with Grounded Entity Biographies"

Disassembly and Analysis of the American Sega Dreamcast VMU BIOS (v1.05)

Tiny 3D Engine for the Sega Dreamcast's Visual Memory Unit by Rockin'-B, written in pure LC86k assembly.

Audio Driver for (dreamcast) VMu

Persistent visual memory for AI agents — capture screenshots, embed with CLIP ViT-B/32, compare, recall. MCP server + Rust core library.

Best flashcard tools, decks, and spaced repetition strategies for memorizing German words and phrases.

LibPerspective is a utility library for writing software on Sega Dreamcast VMU - By Kresna

[ICML 2025] Separating Knowledge with Procedural Data

Incomplete implementation of the Sega Dreamcast VMU in VHDL, based on the ElysianVMU Emulator

MEMO, a visual memory assistant that prioritizes useful uncertainty over confident guesses.

Test your short term memory skills with this simple game. Can you do better than a chimp?

OpenClaw integration for Polymath visual memory through mneme-mcp

Codex插件，能将图片沉淀为可复用的视觉记忆，助你持续创作出美学风格一致的图片。The Codex plugin can turn images into reusable visual memories, helping you consistently create images with a unified aesthetic style.

Visual memo app for every “wow” moment, no need to rack your brain for words.

Compare code reviews across models in Claude Code with ranked todos and subagent dispatch via MCP.

Natural-language robot navigation with multi-agent planning, persistent visual memory, and visual goal verification. Built for ROS 2 and Nav2.

DINOv2-based fixed-budget video memory selection through temporal representation dynamics.

Ctrl+F for the physical world. A camera on a SiMa.ai Modalix that remembers everything in a room and answers out loud: where is it, who took it, what happened. Segmentation + pose + tracking + a vision-language model on one sub-10 W chip. No cloud.