Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386075798> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W4386075798 abstract "Deep neural networks (DNNs) have enabled astounding progress in several vision-based problems. Despite showing high predictive accuracy, recently, several works have revealed that they tend to provide overconfident predictions and thus are poorly calibrated. The majority of the works addressing the miscalibration of DNNs fall under the scope of classification and consider only in-domain predictions. However, there is little to no progress in studying the calibration of DNN-based object detection models, which are central to many vision-based safety-critical applications. In this paper, inspired by the train-time calibration methods, we propose a novel auxiliary loss formulation that explicitly aims to align the class confidence of bounding boxes with the accurateness of predictions (i.e. precision). Since the original formulation of our loss depends on the counts of true positives and false positives in a mini-batch, we develop a differentiable proxy of our loss that can be used during training with other application-specific loss functions. We perform extensive experiments on challenging in-domain and out-domain scenarios with six benchmark datasets including MS-COCO, Cityscapes, Sim10k, and BDD100k. Our results reveal that our train-time loss surpasses strong calibration baselines in reducing calibration error for both in and out-domain scenarios. Our source code and pre-trained models are available at https://github.com/akhtarvision/bpc_calibration" @default.
- W4386075798 created "2023-08-23" @default.
- W4386075798 creator A5000709367 @default.
- W4386075798 creator A5012655817 @default.
- W4386075798 creator A5037593408 @default.
- W4386075798 creator A5079112987 @default.
- W4386075798 date "2023-06-01" @default.
- W4386075798 modified "2023-09-27" @default.
- W4386075798 title "Bridging Precision and Confidence: A Train-Time Loss for Calibrating Object Detection" @default.
- W4386075798 cites W2183341477 @default.
- W4386075798 cites W2194775991 @default.
- W4386075798 cites W2340897893 @default.
- W4386075798 cites W2412782625 @default.
- W4386075798 cites W2736536606 @default.
- W4386075798 cites W2748021867 @default.
- W4386075798 cites W2963351448 @default.
- W4386075798 cites W2963730616 @default.
- W4386075798 cites W2963995504 @default.
- W4386075798 cites W2981207549 @default.
- W4386075798 cites W2982770724 @default.
- W4386075798 cites W2983943451 @default.
- W4386075798 cites W3011367903 @default.
- W4386075798 cites W3021332602 @default.
- W4386075798 cites W3035564946 @default.
- W4386075798 cites W3099111334 @default.
- W4386075798 cites W3175506488 @default.
- W4386075798 cites W4214806325 @default.
- W4386075798 cites W4214893857 @default.
- W4386075798 cites W4312707827 @default.
- W4386075798 cites W4312814827 @default.
- W4386075798 cites W639708223 @default.
- W4386075798 doi "https://doi.org/10.1109/cvpr52729.2023.01104" @default.
- W4386075798 hasPublicationYear "2023" @default.
- W4386075798 type Work @default.
- W4386075798 citedByCount "0" @default.
- W4386075798 crossrefType "proceedings-article" @default.
- W4386075798 hasAuthorship W4386075798A5000709367 @default.
- W4386075798 hasAuthorship W4386075798A5012655817 @default.
- W4386075798 hasAuthorship W4386075798A5037593408 @default.
- W4386075798 hasAuthorship W4386075798A5079112987 @default.
- W4386075798 hasConcept C105795698 @default.
- W4386075798 hasConcept C119857082 @default.
- W4386075798 hasConcept C13280743 @default.
- W4386075798 hasConcept C134306372 @default.
- W4386075798 hasConcept C153180895 @default.
- W4386075798 hasConcept C154945302 @default.
- W4386075798 hasConcept C165838908 @default.
- W4386075798 hasConcept C174348530 @default.
- W4386075798 hasConcept C185798385 @default.
- W4386075798 hasConcept C205649164 @default.
- W4386075798 hasConcept C2776151529 @default.
- W4386075798 hasConcept C2984842247 @default.
- W4386075798 hasConcept C31258907 @default.
- W4386075798 hasConcept C33923547 @default.
- W4386075798 hasConcept C36503486 @default.
- W4386075798 hasConcept C41008148 @default.
- W4386075798 hasConcept C50644808 @default.
- W4386075798 hasConcept C63584917 @default.
- W4386075798 hasConcept C64869954 @default.
- W4386075798 hasConceptScore W4386075798C105795698 @default.
- W4386075798 hasConceptScore W4386075798C119857082 @default.
- W4386075798 hasConceptScore W4386075798C13280743 @default.
- W4386075798 hasConceptScore W4386075798C134306372 @default.
- W4386075798 hasConceptScore W4386075798C153180895 @default.
- W4386075798 hasConceptScore W4386075798C154945302 @default.
- W4386075798 hasConceptScore W4386075798C165838908 @default.
- W4386075798 hasConceptScore W4386075798C174348530 @default.
- W4386075798 hasConceptScore W4386075798C185798385 @default.
- W4386075798 hasConceptScore W4386075798C205649164 @default.
- W4386075798 hasConceptScore W4386075798C2776151529 @default.
- W4386075798 hasConceptScore W4386075798C2984842247 @default.
- W4386075798 hasConceptScore W4386075798C31258907 @default.
- W4386075798 hasConceptScore W4386075798C33923547 @default.
- W4386075798 hasConceptScore W4386075798C36503486 @default.
- W4386075798 hasConceptScore W4386075798C41008148 @default.
- W4386075798 hasConceptScore W4386075798C50644808 @default.
- W4386075798 hasConceptScore W4386075798C63584917 @default.
- W4386075798 hasConceptScore W4386075798C64869954 @default.
- W4386075798 hasLocation W43860757981 @default.
- W4386075798 hasOpenAccess W4386075798 @default.
- W4386075798 hasPrimaryLocation W43860757981 @default.
- W4386075798 hasRelatedWork W2593809839 @default.
- W4386075798 hasRelatedWork W2768865287 @default.
- W4386075798 hasRelatedWork W2801801420 @default.
- W4386075798 hasRelatedWork W2939059980 @default.
- W4386075798 hasRelatedWork W2963037989 @default.
- W4386075798 hasRelatedWork W3014614820 @default.
- W4386075798 hasRelatedWork W3049769619 @default.
- W4386075798 hasRelatedWork W3090145263 @default.
- W4386075798 hasRelatedWork W4225418833 @default.
- W4386075798 hasRelatedWork W4300931582 @default.
- W4386075798 isParatext "false" @default.
- W4386075798 isRetracted "false" @default.
- W4386075798 workType "article" @default.