AI Paper Atlas

Research map / Detection and segmentation / CNN features

Unsupervised Pre-Training for Detection Transformers

Zhigang Dai, Bolun Cai, Yugeng Lin, Junying Chen

TPAMI 2022 · Detection and segmentation › CNN features · approach: Region-based and query-based detectors · 108 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
Unsupervised transformer pre-training (150 epochs)UP-DETR · PASCAL VOCAP: 49.9 -> 56.1Improved
Unsupervised transformer pre-training (300 epochs)UP-DETR · PASCAL VOCAP: 54.1 -> 57.2Improved
Unsupervised pre-training (150 epochs)UP-DETR · COCO val2017AP: 39.7 -> 40.5Improved
Unsupervised pre-training (300 epochs)UP-DETR · COCO val2017AP: 42.1 -> 43.1Improved
Unsupervised pre-training for one-shot detectionUP-DETR · VOC07 one-shot seen classesAP50: 46.2 -> 69.0Improved
Unsupervised pre-training for one-shot detectionUP-DETR · VOC07 one-shot unseen classesAP50: 57.3 -> 73.1Improved
Unsupervised pre-training for panoptic segmentationUP-DETR · COCO panoptic val2017PQ: 44.3 -> 44.7Improved
Use multi-query patches M=10 instead of single-query M=1 (ablation)UP-DETR · PASCAL VOC (ablation)AP: 53.1 -> 54.9Improved
Freeze the CNN backbone during pre-training (ablation)UP-DETR · PASCAL VOC (ablation)AP50: 74.0 -> 78.7Improved
Add feature reconstruction with frozen CNN (ablation)UP-DETR · PASCAL VOC (ablation)AP50: 78.7 -> 78.7No clear effect
Add feature reconstruction without frozen CNN (ablation)UP-DETR · PASCAL VOC (ablation)AP50: 74.0 -> 62.0Hurt
Add attention mask in pre-training (ablation)UP-DETR · PASCAL VOC fine-tuningAPNo clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 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