iiinova — where your digital self keeps evolving

Nikita Araslanov @neekans.bsky.social · 28/05/2026

This yields compelling results on dense vision tasks: video object segmentation, (zero-shot) semantic segmentation and surface normal estimation. Joint work with @masundermeyer.bsky.social, Hidenobu Matsuki, David Joseph Tan and Federico Tombari @munichcenterml.bsky.social

000

Nikita Araslanov @neekans.bsky.social · 28/05/2026

𝗟𝗜𝗟𝗔 trains on videos without manual annotation. Key idea: An optimal linear mapping that predicts dense cues (e.g. depth, flow), estimated on one video frame, should also predict the corresponding cues of another frame from the same video.

100

Nikita Araslanov @neekans.bsky.social · 28/05/2026

In-context learning suggests that a model has learned versatile representations. What if we use in-context learning itself as a training task for visual representations? 📣 Introducing 𝗟𝗜𝗟𝗔: 𝗟𝗶𝗻𝗲𝗮𝗿 𝗜𝗻-𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 ✨ CVPR 2026 Oral ✨ Paper, code and more: lila-pixels.github.io #CVPR2026 🧵👇

100

Nikita Araslanov @neekans.bsky.social · 02/12/2025

📢 NeurIPS 2025 Spotlight 📢 Can we embed motion into image representations? FlowFeat embeds optical flow into pixel-level representations, which results in sharp feature grids, especially for dynamic objects. Project website: tum-vision.github.io/flowfeat With Anna Sonnweber and Daniel Cremers.

062

Reposted by Nikita Araslanov

Dominik Schnaus @schnaus.bsky.social · 03/06/2025

Can we match vision and language representations without any supervision or paired data? Surprisingly, yes! Our #CVPR2025 paper with @neekans.bsky.social and @dcremers.bsky.social shows that the pairwise distances in both modalities are often enough to find correspondences. ⬇️ 1/4

12812

Reposted by Nikita Araslanov

Zhenjun Zhao @ericzzj.bsky.social · 23/04/2025

Back on Track: Bundle Adjustment for Dynamic Scene Reconstruction Weirong Chen, @ganlinzhang.xyz, @fwimbauer.bsky.social, Rui Wang, @neekans.bsky.social, Andrea Vedaldi, @dcremers.bsky.social tl;dr: learning-based 3D point tracker decouples camera and object-based motion arxiv.org/abs/2504.14516

0136

Reposted by Nikita Araslanov

Visual Inference Lab @visinf.bsky.social · 04/04/2025

📢 #CVPR2025 Highlight: Scene-Centric Unsupervised Panoptic Segmentation 🔥 We present CUPS, the first unsupervised panoptic segmentation method trained directly on scene-centric imagery. Using self-supervised features, depth & motion, we achieve SotA results! 🌎 visinf.github.io/cups

1226

Nikita Araslanov @neekans.bsky.social · 04/02/2025

“After carefully reading the other reviews and the author response, I keep my score ‘borderline’” 😣 #CVPR2025

040

Reposted by Nikita Araslanov

Justus Thies @justusthies.bsky.social · 14/01/2025

My group is looking for motivated PhD students that want to work on the future of digital humans. Within the ERC project 'LeMo: Learning Digital Humans in Motion' there are two open positions: www.career.tu-darmstadt.de/HPv3.Jobs/TU... www.career.tu-darmstadt.de/HPv3.Jobs/TU...

0197