AI Paper Atlas

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

Scaling Open-Vocabulary Object Detection

Matthias Minderer, Alexey A. Gritsenko, Neil Houlsby

NeurIPS 2023 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary object detection · 42 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
Self-train on WebLI image-text data using N-gram queriesOWL detector · LVIS val rareAP: 31.2 to 34.9Improved
Fine-tune the self-trained model on LVIS baseOWL-ST CLIP L/14 · LVIS val rareAP: 34.9 to 44.6Improved
Fine-tune the self-trained model on LVIS baseOWL-ST CLIP L/14 · ODinW 13mean AP: 53.0 to 50.1Hurt
Add curated-class queries to the N-gram pseudo-label vocabularyOWL-ST+FT CLIP L/14 · LVIS val rareAP: 44.6 to 45.9Improved
Use machine-generated N-gram queries rather than only a human-curated vocabularyOWL-ST+FT CLIP L/14 · ODinW 13mean AP: 48.7 to 50.1Improved
Average the weights of the self-trained and fine-tuned checkpointsOWL-ST CLIP L/14 · ODinW 13mean AP: 48.7 to 56.3Improved
Drop 0.50 of image tokens by pixel variance (ablation)OWL-ViT L/14 · LVIS val rareAP: 31.8 ± 1.16 to 30.8No clear effect
Drop 0.70 of image tokens by pixel variance (ablation)OWL-ViT L/14 · LVIS val rareAP: 31.8 ± 1.16 to 28.2Hurt
Use a strict pseudo-annotation confidence threshold instead of a moderate threshold (ablation)OWL-ST · LVISAPHurt
Train on curated classes instead of diverse machine-generated text queries (ablation)OWL-ST · unseen classes and ODinWAPHurt
Co-train on WebLI and LVIS rather than average self-trained and fine-tuned weights (ablation)OWL-ST · open-world performanceAPHurt

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