Sophia Sirko-Galouchenko 🇺🇦
@ssirko.bsky.social
PhD student in visual representation learning at Valeo.ai and Sorbonne Université (MLIA)
In our paper DIP, we use DiffCut to generate segmentation pseudo-labels - the masks are very high-fidelity, which greatly boosts supervision quality 👏
November 5, 2025 at 6:42 PM
In our paper DIP, we use DiffCut to generate segmentation pseudo-labels - the masks are very high-fidelity, which greatly boosts supervision quality 👏
Yesterday was PhD defense day for @paulcouairon.bsky.social at @mlia-isir.bsky.social 🎓
His thesis focused on structured visual representations:
• VidEdit - zero-shot text-to-video editing
• DiffCut - zero-shot segmentation via diffusion features
• JAFAR - high-res visual representation upsampling
His thesis focused on structured visual representations:
• VidEdit - zero-shot text-to-video editing
• DiffCut - zero-shot segmentation via diffusion features
• JAFAR - high-res visual representation upsampling
November 5, 2025 at 6:42 PM
Yesterday was PhD defense day for @paulcouairon.bsky.social at @mlia-isir.bsky.social 🎓
His thesis focused on structured visual representations:
• VidEdit - zero-shot text-to-video editing
• DiffCut - zero-shot segmentation via diffusion features
• JAFAR - high-res visual representation upsampling
His thesis focused on structured visual representations:
• VidEdit - zero-shot text-to-video editing
• DiffCut - zero-shot segmentation via diffusion features
• JAFAR - high-res visual representation upsampling
Thrilled to present DIP at #ICCV2025! Great discussions and insightful questions during the poster session. Thank you to everyone who stopped by!
October 24, 2025 at 1:46 AM
Thrilled to present DIP at #ICCV2025! Great discussions and insightful questions during the poster session. Thank you to everyone who stopped by!
6/n Benefits 💪
- < 9h on a single A100 gpu.
- Improves across 6 segmentation benchmarks
- Boosts performance for in-context depth prediction.
- Plug-and-play for different ViTs: DINOv2, CLIP, MAE.
- Robust in low-shot and domain shift.
- < 9h on a single A100 gpu.
- Improves across 6 segmentation benchmarks
- Boosts performance for in-context depth prediction.
- Plug-and-play for different ViTs: DINOv2, CLIP, MAE.
- Robust in low-shot and domain shift.
June 25, 2025 at 7:21 PM
6/n Benefits 💪
- < 9h on a single A100 gpu.
- Improves across 6 segmentation benchmarks
- Boosts performance for in-context depth prediction.
- Plug-and-play for different ViTs: DINOv2, CLIP, MAE.
- Robust in low-shot and domain shift.
- < 9h on a single A100 gpu.
- Improves across 6 segmentation benchmarks
- Boosts performance for in-context depth prediction.
- Plug-and-play for different ViTs: DINOv2, CLIP, MAE.
- Robust in low-shot and domain shift.
5/n Why is DIP unsupervised?
DIP doesn't require manually annotated segmentation masks for its post-training. To accomplish this, it leverages Stable Diffusion (via DiffCut) alongside DINOv2R features to automatically construct in-context pseudo-tasks for its post-training.
DIP doesn't require manually annotated segmentation masks for its post-training. To accomplish this, it leverages Stable Diffusion (via DiffCut) alongside DINOv2R features to automatically construct in-context pseudo-tasks for its post-training.
June 25, 2025 at 7:21 PM
5/n Why is DIP unsupervised?
DIP doesn't require manually annotated segmentation masks for its post-training. To accomplish this, it leverages Stable Diffusion (via DiffCut) alongside DINOv2R features to automatically construct in-context pseudo-tasks for its post-training.
DIP doesn't require manually annotated segmentation masks for its post-training. To accomplish this, it leverages Stable Diffusion (via DiffCut) alongside DINOv2R features to automatically construct in-context pseudo-tasks for its post-training.
4/n Meet Dense In-context Post-training (DIP)! 🔄
- Meta-learning inspired: adopts episodic training principles
- Task-aligned: Explicitly mimics downstream dense in-context tasks during post-training.
- Purpose-built: Optimizes the model for dense in-context performance.
- Meta-learning inspired: adopts episodic training principles
- Task-aligned: Explicitly mimics downstream dense in-context tasks during post-training.
- Purpose-built: Optimizes the model for dense in-context performance.
June 25, 2025 at 7:21 PM
4/n Meet Dense In-context Post-training (DIP)! 🔄
- Meta-learning inspired: adopts episodic training principles
- Task-aligned: Explicitly mimics downstream dense in-context tasks during post-training.
- Purpose-built: Optimizes the model for dense in-context performance.
- Meta-learning inspired: adopts episodic training principles
- Task-aligned: Explicitly mimics downstream dense in-context tasks during post-training.
- Purpose-built: Optimizes the model for dense in-context performance.
2/n What is dense in-context scene understanding?
Formulate dense prediction tasks as nearest-neighbor retrieval problems using patch feature similarities between query and the labeled prompt images (introduced in @ibalazevic.bsky.social et al.’s HummingBird; figure below from their work).
Formulate dense prediction tasks as nearest-neighbor retrieval problems using patch feature similarities between query and the labeled prompt images (introduced in @ibalazevic.bsky.social et al.’s HummingBird; figure below from their work).
June 25, 2025 at 7:21 PM
2/n What is dense in-context scene understanding?
Formulate dense prediction tasks as nearest-neighbor retrieval problems using patch feature similarities between query and the labeled prompt images (introduced in @ibalazevic.bsky.social et al.’s HummingBird; figure below from their work).
Formulate dense prediction tasks as nearest-neighbor retrieval problems using patch feature similarities between query and the labeled prompt images (introduced in @ibalazevic.bsky.social et al.’s HummingBird; figure below from their work).
1/n 🚀New paper out - accepted at #ICCV2025!
Introducing DIP: unsupervised post-training that enhances dense features in pretrained ViTs for dense in-context scene understanding
Below: Low-shot in-context semantic segmentation examples. DIP features outperform DINOv2!
Introducing DIP: unsupervised post-training that enhances dense features in pretrained ViTs for dense in-context scene understanding
Below: Low-shot in-context semantic segmentation examples. DIP features outperform DINOv2!
June 25, 2025 at 7:21 PM
1/n 🚀New paper out - accepted at #ICCV2025!
Introducing DIP: unsupervised post-training that enhances dense features in pretrained ViTs for dense in-context scene understanding
Below: Low-shot in-context semantic segmentation examples. DIP features outperform DINOv2!
Introducing DIP: unsupervised post-training that enhances dense features in pretrained ViTs for dense in-context scene understanding
Below: Low-shot in-context semantic segmentation examples. DIP features outperform DINOv2!