Nicolas Audebert

@nshaud.bsky.social

894 Followers 209 Following

Neuron trainer and Earth Observation mage at IGN France. Kicking stuff until it works. 🌐 https://nicolas.audebert.at 🐘 http://mastodon.social/@nsh (he/him)

"This is fine" dog (brown dog with hat on a stool, surrounded by flames)

Façade d'un bâtiment de 4 étages entièrement vitré (ENPC/Géodata Paris à Champs sur Marne, sur le campus Descartes)

TerraBytes II call for papers announcement
"Towards global datasets and models for Earth Observation"
With organizers from ESA Phi-lab, LASTIG, Le Cnam, KTH, Universidad de Chile, and asterisk labs

Form asking "What's going on?"
- It's annoying or not interesting
- (checked) I'm in this photo and I don't like it
- I think it shouldn't be on Facebook
- It's spam

Please be kindly informed that the goal of TNNLS is to only allow truly high-quality papers (significantly contributing to NN and the related learning systems) to be published by the journal.

*The current acceptance rate of TNNLS is about 18%, which is still too high.*

ChatGPT convinces you to willingly co-operate in this project the same way the bank machine did in my first example: by being really, really useful. ChatGPT is qualitatively different to Claude because it remembers me between chats. It’s formed a ‘theory of mind’ that helps it anticipate my needs and respond in useful ways. Here’s how it explained this new feature to me:

“Because I’ve worked with Inger for a while, I have a rough internal model of her work, audience and preferred voice. When she asks for help, I don’t just generate generic academic advice: I adapt explanations, examples and templates to fit her context – thesis writers, supervisors, time-poor academics – and the way she usually speaks and writes. She still decides what to keep, cut or rewrite, but my suggestions start much closer to what she actually needs than if every interaction began from scratch.”

*highlighted*
It’s true. I treat ChatGPT like an autistic co-worker whose special interest is my work (before you get upset about that statement, I think the special interests side of Autism can be a fantastic academic asset).

Planche de soutenance de thèse :

3D LiDAR point cloud segmentation of small objects

MERIZETTR Maxime
Le 08 octobre 2025

Jury :
Jérôme Verdun, professeur des universités, Cnam-GeF, ESGT, directeur de thèse
Nicolas Audebert, directeur de recherche junior, LASTIG, IGN, co-directeur
Tania Landes, professeure des universités, ICube, INSA, rapportrice
Bruno Vallet, directeur de recherche, LASTIG, IGN, rapporteur
Alexandre Boulch, ingénieur de recherche, Valeo.ai, examinateur
Adrien Gressin, professeur associé, HEIG-VD, examinateur
Pierre Kervella, géomètre expert Quarta, professeur associé, Cnam-GeF ESGT, invité

Digital score in 2/4 with two measures. Below is shown how the score is represented in the state machine, with the various states in the sequence (opening a new bar, playing a note, inserting a rest, starting/ending a chord, etc.).

Christian Igel presenting a slide showing a herd of cows, entitled "Cows from Space: Mapping Livestock in Senegal"

Four pictures 3D chairs flying in the air, over pictural backgrounds (mountains on the top row and an old building on the bottom row). The first two columns are nearly identical images, but the chairs have been slightly moved. The third and last column show the optical flow between the two images, i.e. a color representation of how much every pixel has moved and in which direction.

A textureless mesh of a face on a blue background. On the left, sliders control various facial properties such as "Lower lip", "Upper lip", "Jaw line", "Chin width", "Chin Point", etc.

A night photograph of the Vancouver convention center, near the harbor. The building is illuminated with the waterfront in the foreground.

À gauche :
une ligne de quatre exemples de plan de villes sans aucun bâtiment.
en-dessous, une ligne des mêmes plans mais les empreintes au sol ont été ajoutées.

À droite : succession de cinq images :
1. photo d'un bâtiment vu de biais
2. image de ce bâtiment mais la façade a été segmentée (les fenêtres, portes, murs, colonnes et toits sont dans des couleurs différentes)
3. reconstruction 3D du bâtiment
4. bâtiment différent, généré automatiquement mais dont la structure ressemble au bâtiment 1
5. modèle 3D d'un bâtiment similaire généré automatiquement

Meme de Mourinho qui retire son casque audio avec un air de dégoût

Screenshot of Thunderbird, showing reviewer invitations for WACV 2025, AAAI 2025 and 3DV 2025, all received in the last week.

Un schéma complexe en TikZ avec des gradients de couleur et des flèches reliant une variété en 3D à un carré en 2D.