Matches in SemOpenAlex for { <https://semopenalex.org/work/W2972603178> ?p ?o ?g. }
Showing items 1 to 77 of
77
with 100 items per page.
- W2972603178 abstract "In training deep object detectors, the problem of foreground-background imbalance has been addressed by several heuristic methods, such as online hard example mining (OHEM), Focal Loss, and gradient harmonizing mechanism (GHM). These methods either re-sample the training examples (i.e. hard sampling), or re-weight them discriminatively (i.e. soft sampling). In this paper, we challenge the necessity of such hard/soft sampling heuristics in training deep object detectors. First, without hard/soft sampling, we reveal that the scale and the stability of the classification loss greatly influence the final accuracy. Thus, we propose a guided loss scaling technique to control the classification loss during training, without using any hyper-parameter. We also propose to optimally initialize the model to ensure the stability of the classification loss. Moreover, we propose an adaptive thresholding technique to refine predictions during inference. These three ingredients constitute our Sampling-Free mechanism, which is fully data diagnostic and avoids the laborious hyper-parameter search for hard/soft sampling. We verify the effectiveness of our Sampling-Free mechanism in training one-stage, two-stage, multi-stage, and anchor-free object detectors, where our method always achieves higher accuracy on COCO and PASCAL VOC datasets. We also use the Sampling-Free mechanism for instance segmentation to demonstrate its generalization ability. Code is released at: this https URL." @default.
- W2972603178 created "2019-09-19" @default.
- W2972603178 creator A5000312791 @default.
- W2972603178 creator A5048237545 @default.
- W2972603178 creator A5056036725 @default.
- W2972603178 creator A5082284502 @default.
- W2972603178 date "2019-09-11" @default.
- W2972603178 modified "2023-09-26" @default.
- W2972603178 title "Revisiting Foreground-Background Imbalance in Object Detectors" @default.
- W2972603178 hasPublicationYear "2019" @default.
- W2972603178 type Work @default.
- W2972603178 sameAs 2972603178 @default.
- W2972603178 citedByCount "0" @default.
- W2972603178 crossrefType "posted-content" @default.
- W2972603178 hasAuthorship W2972603178A5000312791 @default.
- W2972603178 hasAuthorship W2972603178A5048237545 @default.
- W2972603178 hasAuthorship W2972603178A5056036725 @default.
- W2972603178 hasAuthorship W2972603178A5082284502 @default.
- W2972603178 hasConcept C111919701 @default.
- W2972603178 hasConcept C112972136 @default.
- W2972603178 hasConcept C115961682 @default.
- W2972603178 hasConcept C119857082 @default.
- W2972603178 hasConcept C127705205 @default.
- W2972603178 hasConcept C140779682 @default.
- W2972603178 hasConcept C153180895 @default.
- W2972603178 hasConcept C154945302 @default.
- W2972603178 hasConcept C191178318 @default.
- W2972603178 hasConcept C199360897 @default.
- W2972603178 hasConcept C2776214188 @default.
- W2972603178 hasConcept C41008148 @default.
- W2972603178 hasConcept C75608658 @default.
- W2972603178 hasConcept C76155785 @default.
- W2972603178 hasConcept C89600930 @default.
- W2972603178 hasConcept C94915269 @default.
- W2972603178 hasConceptScore W2972603178C111919701 @default.
- W2972603178 hasConceptScore W2972603178C112972136 @default.
- W2972603178 hasConceptScore W2972603178C115961682 @default.
- W2972603178 hasConceptScore W2972603178C119857082 @default.
- W2972603178 hasConceptScore W2972603178C127705205 @default.
- W2972603178 hasConceptScore W2972603178C140779682 @default.
- W2972603178 hasConceptScore W2972603178C153180895 @default.
- W2972603178 hasConceptScore W2972603178C154945302 @default.
- W2972603178 hasConceptScore W2972603178C191178318 @default.
- W2972603178 hasConceptScore W2972603178C199360897 @default.
- W2972603178 hasConceptScore W2972603178C2776214188 @default.
- W2972603178 hasConceptScore W2972603178C41008148 @default.
- W2972603178 hasConceptScore W2972603178C75608658 @default.
- W2972603178 hasConceptScore W2972603178C76155785 @default.
- W2972603178 hasConceptScore W2972603178C89600930 @default.
- W2972603178 hasConceptScore W2972603178C94915269 @default.
- W2972603178 hasLocation W29726031781 @default.
- W2972603178 hasOpenAccess W2972603178 @default.
- W2972603178 hasPrimaryLocation W29726031781 @default.
- W2972603178 hasRelatedWork W2345032737 @default.
- W2972603178 hasRelatedWork W2520548719 @default.
- W2972603178 hasRelatedWork W2738982359 @default.
- W2972603178 hasRelatedWork W2773456899 @default.
- W2972603178 hasRelatedWork W2790967904 @default.
- W2972603178 hasRelatedWork W2884132160 @default.
- W2972603178 hasRelatedWork W2898858253 @default.
- W2972603178 hasRelatedWork W2938614126 @default.
- W2972603178 hasRelatedWork W2966913778 @default.
- W2972603178 hasRelatedWork W2970328181 @default.
- W2972603178 hasRelatedWork W2972571929 @default.
- W2972603178 hasRelatedWork W2987591514 @default.
- W2972603178 hasRelatedWork W2990739996 @default.
- W2972603178 hasRelatedWork W3012245929 @default.
- W2972603178 hasRelatedWork W3034826836 @default.
- W2972603178 hasRelatedWork W3036277013 @default.
- W2972603178 hasRelatedWork W3121091114 @default.
- W2972603178 hasRelatedWork W3129499993 @default.
- W2972603178 hasRelatedWork W3161773563 @default.
- W2972603178 hasRelatedWork W2955671438 @default.
- W2972603178 isParatext "false" @default.
- W2972603178 isRetracted "false" @default.
- W2972603178 magId "2972603178" @default.
- W2972603178 workType "article" @default.