Johann Diep
Johann Diep
  • Видео 9
  • Просмотров 2 598
Le CopyChat - Robot, Repeat After Me!
Le CopyChat revolutionizes how robots learn 🤖✨ by enabling them to mimic tasks demonstrated by humans through simple video footage, such as RUclips clips 🎥👩‍🏫.
» GitHub: github.com/johanndiep 💻 ➡️ Code & Projects
» LinkedIn: linkedin.com/in/johann-diep 🌐 ➡️ Professional Network
» Instagram: oxjohann 📸 ➡️ Life & Moments
Просмотров: 65

Видео

Easily Build Your Own AI Robot with Mistral LLM!
Просмотров 120Месяц назад
Robots today are like computers in the 90s-cool but tough to use! To make them part of our daily lives, we need to make sense of their data easily (like drones over crops 🌾) and make controlling them super simple 🚀. Let’s strip away the complexity and bring robots into homes and stores as easy-to-use helpers 🤖! The future is closer than you think 💡! » GitHub: github.com/johanndiep 💻 ➡️ Code & P...
Le Chat Robot - LLM for Robot Trajectory Generation - Mistral Online Fine-tuning Hackathon 2024
Просмотров 1 тыс.3 месяца назад
This is the presentation of the project called "Le Chat Robot" as part of the Mistral Online Fine-tuning Hackathon 2024. Controlling robotic arms requires a profound grasp of concepts like object detection and arm trajectory planning, restricting their use to specialized engineers and researchers. To democratize access to these advanced machines (e.g. for retail workers or farmers), it is impor...
Mobile Cobot LLM Testing - Waveshare UGV Beast with RoArm-M2-S
Просмотров 2673 месяца назад
Simply testing out mobile cobots with LLMs.
Le ChatOn Vision - Visual LLM for Robotics and more - Mistral Hackathon 2024 in Paris
Просмотров 2034 месяца назад
This is the presentation of the project called "Le ChatOn Vision" as part of the Mistral Hackathon 2024 in Paris.
Investigating the Performance of LCNS with Visual-Inertial Odometry for Lunar Rover Navigation
Просмотров 3932 года назад
This video showcases the investigation of possible performance increase using LCNS/GNSS with current visual and inertial techniques used in lunar rover navigation. This project will be presented at the NAVITEC 2022 conference. Authors: Johann Diep, Richard Dennis Swinden, Paolo Zoccarato, Martin Azkarate Vecilla, Pietro Giordano, Javier Ventura-Traveset and Paolo Crosta Commercial User Segment ...
Performance Analysis of Mass-Market GNSS Receivers in UAV Applications
Просмотров 1382 года назад
This video showcases the assessment of the performance of various GNSS technologies in aeronautical drone maneuvers. The benchmarking is presented in the paper: "Performance Analysis of Mass-Market GNSS Receivers in UAV Applications", NAVITEC 2022. Authors: Johann Diep, David Gómez-Casco, Xurxo Otero Villamide, Richard Dennis Swinden and Paolo Crosta Commercial User Segment and Navigation Syste...
UAV Trajectory Following with Uncorrected UWB Measurement Model
Просмотров 1454 года назад
This video showcase erroneous localization with commercially available UWB technology. An improvement is demonstrated in the following video: ruclips.net/video/YJo8tMqly8E/видео.html This project was conducted in the context of a master thesis at ETH Zurich: gitlab.com/jdiep/master-thesis Author: johanndiep@gmail.com
UAV Trajectory Following with Corrected UWB Measurement Model
Просмотров 2294 года назад
This video showcase the correction of the localization with commercially available UWB technology. Compare this outcome with the results of uncorrected localization: ruclips.net/video/BqnEMfDYWC4/видео.html This project was conducted in the context of a master thesis at ETH Zurich: gitlab.com/jdiep/master-thesis Author: johanndiep@gmail.com

Комментарии

  • @BobRos2
    @BobRos2 Месяц назад

    This was the most interesting video in the Mistral office hour yesterday, thanks!

  • @swannschilling474
    @swannschilling474 3 месяца назад

    This is great, I would really like to try this on my robot arm!!

  • @rebotnixai
    @rebotnixai 3 месяца назад

    Does the robot then have to have a functioning internet connection in order to receive the commands from the mistral model via the API interface? that would be a very high security risk for such applications as well as a high risk for stable operation if the internet connection does not exist or there is a fault in the api from the mixtral backend. thank you very much for your answer.

    • @oxjohann
      @oxjohann 3 месяца назад

      You are correct that many robotic applications involve the collection of sensitive data, necessitating a solution without internet connectivity. Consequently, we opted to build our solution around Mistral's smallest base model. The objective of this project was to demonstrate that even a small model, when sufficiently fine-tuned, can be utilized for complex robotic planning tasks. This approach enables us to host the model directly on an edge computer, ideally the same one that powers the robotic arm, ensuring all computations are performed on-board. This configuration provides a secure and efficient solution.

    • @rebotnixai
      @rebotnixai 3 месяца назад

      @@oxjohann That's exactly what I imagined. I am also developing an edge board for robot control based on NVIDIA JETSON Orin NX and at the end of the year with the new Blackwell and nvidia goort platform. I would be delighted if we could continue to exchange ideas. Is there already a Discord or other channels where like-minded people meet?

  • @oxjohann
    @oxjohann 3 месяца назад

    A mistake was found at 1:01, for both the visual segmentation and the trajectory control, the input should be "search for pen" and "bring me an pen" respectively.

  • @ArghyaChatterjeeJony
    @ArghyaChatterjeeJony 2 года назад

    Hi, thanks for sharing your sharing your research work. I have a question. I can't see the SPARTAN github repo. Is it opensource? If so, can you give me the link to the repo?

    • @oxjohann
      @oxjohann 2 года назад

      Thanks for the comment, to my knowledge the software is to this date only available ESA-internally.

    • @ArghyaChatterjeeJony
      @ArghyaChatterjeeJony 2 года назад

      @@oxjohann Ok. Thanks for letting me know.