STORM: Nvidia paper to directly regress 3D Gaussians + velocities from AV sensor data. Instead of fitting 3D Gaussians to a scene (~30 mins), reconstructs in one forward pass (~200ms).
04.01.2025 04:50 — 👍 29 🔁 6 💬 1 📌 1@olayatu.bsky.social
PhD in deep learning algorithms for underwater visual localization. Things I like: robots, art, birds. Things i don't like: social media taken over by bots.
STORM: Nvidia paper to directly regress 3D Gaussians + velocities from AV sensor data. Instead of fitting 3D Gaussians to a scene (~30 mins), reconstructs in one forward pass (~200ms).
04.01.2025 04:50 — 👍 29 🔁 6 💬 1 📌 1"The 3D Gaussian Splatting Adventure: Past, Present, Future" by George Drettakis is a great lecture. www.youtube.com/watch?v=DjOq...
04.02.2025 19:38 — 👍 18 🔁 7 💬 1 📌 0La próxima pa cuándo? 😃😃😃
08.02.2025 15:07 — 👍 1 🔁 0 💬 0 📌 0Me encanta la falta de ortografía en la plantilla de desmotivaciones
08.02.2025 15:06 — 👍 0 🔁 0 💬 0 📌 0Alguien se acuerda del chonizador? He aplicado mi vasto conocimiento en diseño gráfico y web para hacer una réplica. Adjunto foto de la página original allá por el 2010. La réplica está al abrigo de mi dominio en GitHub: olayasturias.github.io/chonizador/ Conste que está aún en versión alpha.
08.02.2025 15:05 — 👍 0 🔁 0 💬 1 📌 0💡 What we do:
- Survey SLAM algorithms across geometry-based and learning-based frameworks.
- Provide a unified formulation of the SLAM pipeline used across implementations.
- Classify and evaluate SLAM methods based on different environmental conditions.
🔍 Why this matters:
SLAM algorithms face diverse environmental challenges:
- Dynamic elements in outdoor scenes 🌳
- Harsh imaging in underwater environments 🌊
- Blurriness in high-speed setups 🚗💨
Understanding these challenges is crucial for real-world applications.
I. My survey paper: "Monocular visual simultaneous localization and mapping:(r)evolution from geometry to deep learning-based pipelines"
This paper served me as a study of vSLAM and laid the foundations of my research. Maybe it can help you too!
📄 Paper and code: paperswithcode.com/paper/monocu...
In the past, Twitter was a great resource for me to find interesting papers and/or repositories, especially from the ML community. Let's tune my algorithm for Bluesky while I share a bit of my work! 😄
10.12.2024 09:09 — 👍 0 🔁 0 💬 0 📌 0Oleeeeee, doctor!!!!!
19.11.2024 06:41 — 👍 1 🔁 0 💬 0 📌 0