Tetiana Martyniuk
@t-martyniuk.bsky.social
300 followers 65 following 39 posts
a Ukrainian computer vision researcher escaping Twitter madness 😅 3D Computer Vision & Generative modelling PhD student @ Inria Paris & @valeoai.bsky.social | Researcher @ Ukrainian Catholic University https://t-martyniuk.github.io/
Posts Media Videos Starter Packs
t-martyniuk.bsky.social
Some conclusions and food for thought for the future research.
t-martyniuk.bsky.social
Another great event for @valeoai.bsky.social team: a PhD defense of Corentin Sautier.

His thesis «Learning Actionable LiDAR Representations w/o Annotations» covers the papers BEVContrast (learning self-sup LiDAR features), SLidR, ScaLR (distillation), UNIT and Alpine (solving tasks w/o labels).
t-martyniuk.bsky.social
Some insights from soon-to-be Dr. Michele: Conclusions and Perspectives.
t-martyniuk.bsky.social
So excited to attend the PhD defense of @bjoernmichele.bsky.social at @valeoai.bsky.social! He’s presenting his research results of the last 3 years in 3D domain adaptation: SALUDA (unsupervised DA), MuDDoS (multimodal UDA), TTYD (source-free UDA).
t-martyniuk.bsky.social
I took my parents to see it, my mom even cried! She liked it a lot.

I liked it but with a small shortcoming that it was a bit like a fairytale: with his driving style he’d get 12 penalty points in the super license easily within those 9 races 🙈
t-martyniuk.bsky.social
LiDPM is a collaboration between Inria and @valeoai.bsky.social, couldn't have done it without the support of my amazing supervisors @alexandreboulch.bsky.social, @gillespuy.bsky.social, Renaud Marlet and Raoul de Charette!
t-martyniuk.bsky.social
We show that "vanilla" diffusion can be applied to large-scale (here - lidar) non-normalized point clouds.

LiDPM
● does not require any additional regularization,
● inherently extends to unconditional scene generation,
● outperforms "local" diffusion for scene completion on SemanticKITTI.
t-martyniuk.bsky.social
We just released the code of #LiDPM, go ahead and play with it (and don't forget to star 🤭🤩)!

Training and inference code available, along with the model checkpoint.

Github repo: github.com/astra-vision...

#IV2025
t-martyniuk.bsky.social
It’s indeed a particular way to visualize the data :)
t-martyniuk.bsky.social
One author with 14 submissions 🤯

#iv2025
t-martyniuk.bsky.social
30% of the papers were coming from Germany 🤯

France - 5%

#iv2025
t-martyniuk.bsky.social
Some #IV2025 statistics:

1/3 of all submissions were uploaded on the last day 🙃
t-martyniuk.bsky.social
On one hand you might think “The conference destinations that are not exotic enough are not as cool”.

But let me just tell you: I feel like I’m at a wedding in Bukovyna rn, and I mean it in the best way possible! Such an amazing reception and beautiful Romanian folk music :) #iv2025
t-martyniuk.bsky.social
And a really great poster session afterwards! So many relevant questions, I was thrilled to share the details of our work, thanks for stopping by. #iv2025
t-martyniuk.bsky.social
Okay that was stressful 🥲
t-martyniuk.bsky.social
Presenting our project #LiDPM in the afternoon oral session at #IV2025!

Project page: astra-vision.github.io/LiDPM/

w/ @gillespuy.bsky.social, @alexandreboulch.bsky.social, Renaud Marlet, Raoul de Charette

Also, see our poster at 3pm in the Caravaggio room and AMA 😉
t-martyniuk.bsky.social
Multi-sensor fusion review paper presented by Dr. Ziran Wang.

#IV2025
t-martyniuk.bsky.social
Dr. Ignacio Alvarez overviewing Generative AI for synthetic scenario generations, automated HD mapping, behavior of traffic agents, and world modeling.

#IV2025
t-martyniuk.bsky.social
Attending #DLIVA - a Workshop on Data-Driven Learning for Intelligent Vehicle Applications: Foundation Models, Multimodal Fusion and Cooperative Systems.

#IV2025
t-martyniuk.bsky.social
Day 1 of #IV2025 in Cluj-Napoca - let’s go!

Today is the Workshops day.

#autonomousDriving #autonomousVehicles #computerVision #AI
t-martyniuk.bsky.social
Presenting LiDPM today at #CVPRinParis, such a great poster session, many new acquaintances and curious questions! #cvpr #cvpr2025

Next stop: #IV2025.
t-martyniuk.bsky.social
Our paper "LiDPM: Rethinking Point Diffusion for Lidar Scene Completion" got accepted to IEEE IV 2025!

tldr: LiDPM enables high-quality LiDAR completion by applying a vanilla DDPM with tailored initialization, avoiding local diffusion approximations.

Project page: astra-vision.github.io/LiDPM/
t-martyniuk.bsky.social
there should be or*

I don’t launch training on inference (what would that even mean?), but you can definitely see that I lack some sleep 😵‍💫