AI Paper Atlas

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

CLIFF: Continual Latent Diffusion for Open-Vocabulary Object Detection

Wuyang Li, Xinyu Liu, Jiayi Ma, Yixuan Yuan

ECCV 2024 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary object detection · 11 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
add object-to-text diffusion (ablation)Faster RCNN C4 baseline · COCO OVD (1x)mAPn: 32.2 -> 36.6Improved
add object-to-image diffusion (full CDM) (ablation)CLIFF o2t · COCO OVD (1x)mAPn: 36.6 -> 39.5Improved
add Variational Latent Sampler (object-centric noise) (ablation)CLIFF CDM · COCO OVD (1x)mAPn: 39.5 -> 40.0Improved
continual diffusion instead of combinatorial shared diffusion (ablation)CLIFF · COCO OVD (2x)mAPn: 39.9 -> 41.4Improved
use object-centric noise instead of random noise throughout (ablation)CLIFF o2t · COCO OVD (2x)mAPn: 39.2 -> 41.3Improved
add discriminative term to object-to-text loss (ablation)CLIFF · COCO OVD (2x)mAPn: 38.2 -> 41.3Improved
add generative term to object-to-text loss (ablation)CLIFF · COCO OVD (2x)mAPn: 40.1 -> 41.3Improved
increase o2i diffusion steps L from 3 to 5/7/9 (ablation)CLIFF · COCO OVD (2x)mAPn: 41.3 -> 41.3/41.2/41.4No clear effect
reduce o2i steps to L=1 (ablation)CLIFF · COCO OVD (2x)mAPn: 41.3 -> 40.2Hurt
raise o2t steps T from 10 to 20 (ablation)CLIFF · COCO OVD (2x)mAPn: 41.3 -> 39.7Hurt
attentive condition refinement instead of xobj only (ablation)CLIFF o2t condition · COCO OVD (2x)mAPn: 40.4 -> 41.3Improved
CLIFF vs CoDet on overall LVIS AP (CenterNet v2)CLIFF · LVIS (CenterNet v2)AP: 28.2 vs 30.7Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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