ViTacFormer is a unified visuo-tactile framework for dexterous robot manipulation. It fuses high-res visual+tactile data using cross-attention and predicts future tactile signals via an autoregressive head, enabling multi-fingered hands to perform precise, long-horizon tasks.

More details in this thread:

Scan any documents, convert images into text, PDF files, etc. 👍

Whenever I see these posts now it's always depressing how sterile and perfect and not real world at all the test environment needs to be. We have come almost nowhere in years.

Edward scissor hands looking bot 🤖

I'm losing faith in the general-purpose humanoid promise. Let's face it, it's going to be a decade (at least) before the average person could trust one of these things in their kitchen to prepare a meal. Alternatively, bot buddies could be available in less than half of that time

Robots that see and feel ViTacFormer is a big step toward true human-like precision in manipulation.

Merging sight and touch — this is a major leap toward truly human-like robotic dexterity.