Segment Anything without Supervision Exceeds SAM’s AR by over 6.7% and AP by 3.9% using only 1% of labels repo: github.com/frank-xwang/Un… abs: arxiv.org/abs/2406.20081

Announcing one of the most unique and ambitious robot competitions ever: 🏁The Earth Rover Challenge at #IROS2024! ✅Globally distributed in-the-wild evaluation ✅Real world navigation task settings ✅Large training dataset provided Details below 👇

Congrats to ahmad khaliq for his accepted #ECCV2024 paper!!! 𝐕𝐋𝐀𝐃-𝐁𝐮𝐅𝐅: 𝐁𝐮𝐫𝐬𝐭-𝐚𝐰𝐚𝐫𝐞 𝐅𝐚𝐬𝐭 𝐅𝐞𝐚𝐭𝐮𝐫𝐞 𝐀𝐠𝐠𝐫𝐞𝐠𝐚𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐕𝐢𝐬𝐮𝐚𝐥 𝐏𝐥𝐚𝐜𝐞 𝐑𝐞𝐜𝐨𝐠𝐧𝐢𝐭𝐢𝐨𝐧 w. Sourav Garg lead supervisor + co-authors Ming Xu Stephen Hausler

🚀 We present AnyFeature-VSLAM at #RSS2024 next week! 📝 AnyFeature-VSLAM: Automating the Usage of Any Chosen Feature into VSLAM, Alejandro Fontán, Javier Civera, Michael Milford FTSE 🤖🚘🐀🧠📘🗣️ 📅 Session 8. Perception and Navigation 🗓️ 17-Jul ⏰ 10:00-11:00am Paper: roboticsconference.org/program/papers… [1/3]

📌How can we scale imitation learning to long-horizon, multi-object tasks & generalise (spatial & intra-category) from only 10 demos? ✨Introducing Affordance-Centric Policy Decomposition 💡Relative task-frame diffusion policies that self-chain for seamless operation 🧵👇

Baffling European Conference on Computer Vision #ECCV2024 policy: Students cannot register their own paper, and must either rely on a non-student co-author's full registration, or must pay more money. Is the idea that we should be taxing students who write papers without faculty/industry support? Seems bad.

🚀 Excited to release OpenVPRLab! 🎉 An open-source framework for Visual Place Recognition (VPR), featuring extensible, modular, and scalable components, enabling researchers to train/develop deep VPR models with reproducible SOTA performance. 🔗github.com/amaralibey/Ope… 🧵👇

Now a shoutout to Sourav Garg's #SeqMatchNet work for #VPR (an oral at CoRL2021 not 2022) Somayeh Tobias Fischer

Super proud of Somayeh for giving a fantastic final #PhD seminar on #spiking #networks 🧠for #place #recognition 🗺️🧭. Read about her great work 📝 QUTRobotics with myself and Michael Milford FTSE 🤖🚘🐀🧠📘🗣️: scholar.google.com.au/citations?hl=e… 🤖

To help explain the weirdness of LLM Tokenization I thought it could be amusing to translate every token to a unique emoji. This is a lot closer to truth - each token is basically its own little hieroglyph and the LLM has to learn (from scratch) what it all means based on

The question of whether LLMs can reason is, in many ways, the wrong question. The more interesting question is whether they are limited to memorization / interpolative retrieval, or whether they can adapt to novelty beyond what they know. (They can't, at least until you start

We created some visualisations of MASt3R on the #MapFreeReloc dataset. The numbers do not lie, the results are amazing. The public version is a bit worse than the private one which is top on the leaderboard. 0.751 AUC instead of 0.817. Examples in 🧵including public vs private.

Left: Hierarchical model based RL with a large-scale pre-trained world model, auxiliary tasks and skill-discovery and a model for inverse kinematics. Right: PID

7/8 Finally, had to try Transformers (wink-wink)! Several point prompts but all before the transition through the underpass! Roughly - Point prompts for Bumblebee on frame 050, on the Decepticon on frame 089 (before it disintegrates!), a couple on the bridge on frame 175.