AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

Unsupervised Image Representation Learning with Deep Latent Particles

Tal Daniel, Aviv Tamar

ICML 2022 · Detection and segmentation › Open-set and few-shot detection · approach: Unsupervised object and part discovery · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use DLP with 30 particles instead of KeyNet with 30 keypointsUnsupervised landmark regression · MAFLMSE (%): 2.58 -> 2.56Improved
Use DLP with 50 particles instead of KeyNet with 50 keypointsUnsupervised landmark regression · MAFLMSE (%): 2.54 -> 2.43Improved
Add log-variance to the regression features (ablation)DLP with 25 particles · MAFLMSE (%): 2.87 -> 2.52Improved
Add log-variance to the regression features (ablation)DLP with 30 particles · MAFLMSE (%): 2.56 -> 2.49Improved
Add learned particle features to the regression features (ablation)DLP with 30 particles · MAFLMSE (%): 2.56 -> 2.36Improved
Add learned particle features to the regression features (ablation)DLP with 50 particles · MAFLMSE (%): 2.43 -> 2.39Hurt
Use Chamfer-KL instead of Standard KL (ablation)DLP with constant Gaussian prior · MAFLMSE (%): 3.18 -> 3.07Improved
Use Chamfer-KL instead of Standard KL (ablation)DLP with uniform prior proposals · MAFLMSE (%): 3.39 -> 2.99Improved
Use Chamfer-KL instead of Standard KL (ablation)DLP with SSM prior · MAFLMSE (%): 3.24 -> 2.65Improved
Use the SSM prior instead of a constant Gaussian prior (ablation)DLP with Chamfer-KL · MAFLMSE (%): 3.07 -> 2.65Improved
Use the SSM prior instead of uniform prior proposals (ablation)DLP with Chamfer-KL · MAFLMSE (%): 2.99 -> 2.65Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy