Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386075772> ?p ?o ?g. }
- W4386075772 abstract "In this study, we dive deep into the inconsistency of pseudo targets in semi-supervised object detection (SSOD). Our core observation is that the oscillating pseudo-targets undermine the training of an accurate detector. It injects noise into the student's training, leading to severe overfitting problems. Therefore, we propose a systematic solution, termed Consistent-Teacher, to reduce the inconsistency. First, adaptive anchor assignment (ASA) substitutes the static IoU-based strategy, which enables the student network to be resistant to noisy pseudo-bounding boxes. Then we calibrate the subtask predictions by designing a 3D feature alignment module (FAM-3D). It allows each classification feature to adaptively query the optimal feature vector for the regression task at arbitrary scales and locations. Lastly, a Gaussian Mixture Model (GMM) dynamically revises the score threshold of pseudo-bboxes, which stabilizes the number of ground truths at an early stage and remedies the unreliable supervision signal during training. Consistent-Teacher provides strong results on a large range of SSOD evaluations. It achieves 40.0 mAP with ResNet-50 backbone given only 10% of annotated MS-COCO data, which surpasses previous base-lines using pseudo labels by around 3 mAP. When trained on fully annotated MS-COCO with additional unlabeled data, the performance further increases to 47.7 mAP. Our code is available at https://github.com/Adamdad/ConsistentTeacher." @default.
- W4386075772 created "2023-08-23" @default.
- W4386075772 creator A5005222486 @default.
- W4386075772 creator A5021622922 @default.
- W4386075772 creator A5030261532 @default.
- W4386075772 creator A5039785210 @default.
- W4386075772 creator A5057842914 @default.
- W4386075772 creator A5059138390 @default.
- W4386075772 creator A5067731176 @default.
- W4386075772 creator A5074339191 @default.
- W4386075772 creator A5080003280 @default.
- W4386075772 date "2023-06-01" @default.
- W4386075772 modified "2023-09-25" @default.
- W4386075772 title "Consistent-Teacher: Towards Reducing Inconsistent Pseudo-Targets in Semi-Supervised Object Detection" @default.
- W4386075772 cites W2108598243 @default.
- W4386075772 cites W2565639579 @default.
- W4386075772 cites W2570343428 @default.
- W4386075772 cites W2962766617 @default.
- W4386075772 cites W2963037989 @default.
- W4386075772 cites W2963351448 @default.
- W4386075772 cites W2963857746 @default.
- W4386075772 cites W2964241181 @default.
- W4386075772 cites W2982770724 @default.
- W4386075772 cites W3035160371 @default.
- W4386075772 cites W3035395201 @default.
- W4386075772 cites W3035396860 @default.
- W4386075772 cites W3091802947 @default.
- W4386075772 cites W3102701618 @default.
- W4386075772 cites W3158661000 @default.
- W4386075772 cites W3167308647 @default.
- W4386075772 cites W3172507542 @default.
- W4386075772 cites W3176376875 @default.
- W4386075772 cites W3178291178 @default.
- W4386075772 cites W3179069071 @default.
- W4386075772 cites W4214507171 @default.
- W4386075772 cites W4312599584 @default.
- W4386075772 cites W4312605608 @default.
- W4386075772 cites W4312769060 @default.
- W4386075772 cites W4312887059 @default.
- W4386075772 doi "https://doi.org/10.1109/cvpr52729.2023.00316" @default.
- W4386075772 hasPublicationYear "2023" @default.
- W4386075772 type Work @default.
- W4386075772 citedByCount "0" @default.
- W4386075772 crossrefType "proceedings-article" @default.
- W4386075772 hasAuthorship W4386075772A5005222486 @default.
- W4386075772 hasAuthorship W4386075772A5021622922 @default.
- W4386075772 hasAuthorship W4386075772A5030261532 @default.
- W4386075772 hasAuthorship W4386075772A5039785210 @default.
- W4386075772 hasAuthorship W4386075772A5057842914 @default.
- W4386075772 hasAuthorship W4386075772A5059138390 @default.
- W4386075772 hasAuthorship W4386075772A5067731176 @default.
- W4386075772 hasAuthorship W4386075772A5074339191 @default.
- W4386075772 hasAuthorship W4386075772A5080003280 @default.
- W4386075772 hasConcept C115961682 @default.
- W4386075772 hasConcept C119857082 @default.
- W4386075772 hasConcept C138885662 @default.
- W4386075772 hasConcept C153180895 @default.
- W4386075772 hasConcept C154945302 @default.
- W4386075772 hasConcept C162324750 @default.
- W4386075772 hasConcept C177264268 @default.
- W4386075772 hasConcept C187736073 @default.
- W4386075772 hasConcept C199360897 @default.
- W4386075772 hasConcept C22019652 @default.
- W4386075772 hasConcept C2776151529 @default.
- W4386075772 hasConcept C2776401178 @default.
- W4386075772 hasConcept C2776760102 @default.
- W4386075772 hasConcept C2780451532 @default.
- W4386075772 hasConcept C2781238097 @default.
- W4386075772 hasConcept C41008148 @default.
- W4386075772 hasConcept C41895202 @default.
- W4386075772 hasConcept C50644808 @default.
- W4386075772 hasConcept C63584917 @default.
- W4386075772 hasConcept C76155785 @default.
- W4386075772 hasConcept C83665646 @default.
- W4386075772 hasConcept C94915269 @default.
- W4386075772 hasConcept C99498987 @default.
- W4386075772 hasConceptScore W4386075772C115961682 @default.
- W4386075772 hasConceptScore W4386075772C119857082 @default.
- W4386075772 hasConceptScore W4386075772C138885662 @default.
- W4386075772 hasConceptScore W4386075772C153180895 @default.
- W4386075772 hasConceptScore W4386075772C154945302 @default.
- W4386075772 hasConceptScore W4386075772C162324750 @default.
- W4386075772 hasConceptScore W4386075772C177264268 @default.
- W4386075772 hasConceptScore W4386075772C187736073 @default.
- W4386075772 hasConceptScore W4386075772C199360897 @default.
- W4386075772 hasConceptScore W4386075772C22019652 @default.
- W4386075772 hasConceptScore W4386075772C2776151529 @default.
- W4386075772 hasConceptScore W4386075772C2776401178 @default.
- W4386075772 hasConceptScore W4386075772C2776760102 @default.
- W4386075772 hasConceptScore W4386075772C2780451532 @default.
- W4386075772 hasConceptScore W4386075772C2781238097 @default.
- W4386075772 hasConceptScore W4386075772C41008148 @default.
- W4386075772 hasConceptScore W4386075772C41895202 @default.
- W4386075772 hasConceptScore W4386075772C50644808 @default.
- W4386075772 hasConceptScore W4386075772C63584917 @default.
- W4386075772 hasConceptScore W4386075772C76155785 @default.
- W4386075772 hasConceptScore W4386075772C83665646 @default.
- W4386075772 hasConceptScore W4386075772C94915269 @default.
- W4386075772 hasConceptScore W4386075772C99498987 @default.
- W4386075772 hasLocation W43860757721 @default.