DreamDojo: Scaling Robot World Models with 44,000+ Hours of Egocentric Human Video
TL;DR DreamDojo pretrains a robot world model on 44,711 hours of egocentric human video using continuous latent actions as proxy
Read more
Luigi Freda
Computer Vision, AI, Robotics Engineer, PhD
Machine Learning
Probing the 3D Awareness of Visual Foundation Models
As visual foundation models increasingly find their way into SLAM and 3D reconstruction pipelines, an important question arises: are these
Read more
Waymo has unveiled the Waymo World Model
Waymo has unveiled the Waymo World Model, a next-generation simulation platform built on Google DeepMind’s Genie 3. This frontier generative
Read more
Genie 3 – AI That Builds Worlds from Words
Genie 3, developed by Google DeepMind, represents a significant step toward general-purpose world models, which are AI systems that learn the
Read more
A new release of pySLAM is here. Loop-closing, volumetric integration, depth prediction and more improvements.
🎉 pySLAM v2.2.5 is here! This new release includes a visual SLAM pipeline for monocular, stereo, and RGBD cameras with:
Read more
My new pySLAM v2 is out!
I am excited to release pySLAM v2. The new version allows you to play with SLAM techniques, visual-odometry, keyframes, bundle-adjustment,
Read more
Hottest Deep Learning Frameworks 2018
In this very nice article, Jeff compares all the hottest deep learning frameworks that are used in the Academic and
Read more
CNN-SLAM: Real-time dense monocular SLAM with learned depth prediction
In this page, the Authors K. Tateno, F. Tombari, I. Laina and N. Navab present the following paper in which CNNs are used
Read more
Hand-Crafted Features or Learned Local Features?
In this post, Torsten Sattler presents his upcoming CVPR 2017 paper “Comparative Evaluation of Hand-Crafted and Learned Local Features”
Read more
Optimization – why momentum really works
This post explains why momentum works in optimization. Nice interactive images are used to this aim. Here you can find another
Read more
Driverless cars that ‘see’ the road ahead
Micheal Milford explains how to make a driverless car see the road ahead. From his Google+ post: “We were asked to
Read more
Do we need self-driving cars?
This post reports that a self-driving Uber car was involved in a high-speed crash in Tempe, Arizona. No one was
Read more
Bias and variance tradeoff
This very nice post clearly explains the bias and variance tradeoff http://www.learnopencv.com/bias-variance-tradeoff-in-machine-learning/ This tradeoff can be mathematically stated by the
Read more
ElasticFusion – Real-time Light Source Detection
A new journal paper about ElasticFusion has come out from Davison’s group. New exciting results on 3D light source mapping are presented.
Read more
Probabilistic Deep Network
This post explains a trick which allows us to convert neural network outputs into probabilities, with no cost to performance,
Read more