Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
M1 Pro Robot Motion Control
This talk demonstrates how 2D webcam footage is converted into 3D movements to control a humanoid robot using computer vision and pose estimation on an M1 Pro laptop.
Ever wondered if you could make a robot mirror your every move using just your laptop camera? Get ready to see it happen! In this talk, we’ll dive into how simple 2D webcam footage can be transformed into precise 3D movements that a humanoid robot can follow. The fascinating part? This isn’t science fiction – it’s working right now using computer vision and pose estimation techniques. We’ll demonstrate the entire pipeline, from capturing human motion to making the robot move, all running smoothly on a standard M1 Pro laptop. Come explore how we bridged the gap between 2D imagery and 3D robotics, making complex motion control accessible without specialized hardware.
- STCFormerSTCFormer (Spatio-Temporal Criss-cross Transformer) is a high-efficiency model for 3D Human Pose Estimation (HPE), utilizing a decomposed attention mechanism to minimize quadratic computational cost.This is the Spatio-Temporal Criss-cross Transformer: a robust architecture for 3D Human Pose Estimation. STCFormer addresses the quadratic computational cost of full spatio-temporal attention by introducing the STC block, which efficiently decomposes correlation learning into parallel spatial and temporal pathways. The system integrates a Structure-enhanced Positional Embedding (SPE) to factor in explicit human body structure, boosting accuracy. Validated on major benchmarks, the model delivered a state-of-the-art 40.5mm P1 error on the challenging Human3.6M dataset, confirming its superior performance and economic design: it achieves this with significantly fewer parameters than prior state-of-the-art techniques.
- Genesis SimulationGenesis is an ultra-fast, 100% Python generative physics engine (GPU-accelerated) for high-fidelity simulation in robotics and embodied AI.Genesis is a universal physics platform, developed by a consortium of over 20 research labs, built to accelerate robotics and embodied AI data creation. The engine delivers unprecedented speed: it runs simulations up to 430,000 times faster than real-time (e.g., a Franka arm scene) and outperforms conventional GPU simulators like NVIDIA's Isaac Gym by a factor of 10–80x. Built entirely in Python, Genesis unifies multiple physics solvers (Rigid Body, SPH, FEM) and integrates a generative framework: users can prompt the system with natural language to autonomously create complex 4D dynamic worlds, set reward functions, and generate robotic policies. This capability minimizes manual data collection, making Genesis a paradigm shift for synthetic data generation and rapid iteration.
Related projects
NanoBrowser: Building Open-Source Alternative to OpenAI "Operator"
Singapore
This talk demonstrates NanoBrowser, an open-source AI-powered browser extension for web automation, highlighting community contributions, practical use cases,…
BARK Robot Dog
Lausanne
The talk covers the hardware development of BARK, a quadruped robot dog, and explores applying reinforcement learning to…
Building my own ai-sdk
Singapore
Learn how to build a custom AI SDK for multi‑LLM integration, handling provider-specific quirks and edge cases beyond…
AI or DIE: Making an AI game show
Singapore
A walkthrough of building an AI‑driven game show where contestants persuade an AI judge, covering design, implementation, and…
Camera-Based Travel Assistant
Singapore
Learn how to build a camera‑based travel assistant web app with Next.js route handlers, Vercel AI SDK streaming,…
Oneservice Hotline
Singapore
We’ll cover building a multimodal speech‑to‑speech assistant with OpenAI’s Realtime API and function calling, targeting elderly users speaking…