FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3217158093> ?p ?o ?g. }

• W3217158093 abstract "A core component of the recent success of self-supervised learning is cropping data augmentation, which selects sub-regions of an image to be used as positive views in the self-supervised loss. The underlying assumption is that randomly cropped and resized regions of a given image share information about the objects of interest, which the learned representation will capture. This assumption is mostly satisfied in datasets such as ImageNet where there is a large, centered object, which is highly likely to be present in random crops of the full image. However, in other datasets such as OpenImages or COCO, which are more representative of real world uncurated data, there are typically multiple small objects in an image. In this work, we show that self-supervised learning based on the usual random cropping performs poorly on such datasets. We propose replacing one or both of the random crops with crops obtained from an object proposal algorithm. This encourages the model to learn both object and scene level semantic representations. Using this approach, which we call object-aware cropping, results in significant improvements over scene cropping on classification and object detection benchmarks. For example, on OpenImages, our approach achieves an improvement of 8.8% mAP over random scene-level cropping using MoCo-v2 based pre-training. We also show significant improvements on COCO and PASCAL-VOC object detection and segmentation tasks over the state-of-the-art self-supervised learning approaches. Our approach is efficient, simple and general, and can be used in most existing contrastive and non-contrastive self-supervised learning frameworks." @default.
• W3217158093 created "2021-12-06" @default.
• W3217158093 creator A5009078067 @default.
• W3217158093 creator A5015390901 @default.
• W3217158093 creator A5021971352 @default.
• W3217158093 creator A5023195442 @default.
• W3217158093 creator A5037733040 @default.
• W3217158093 creator A5059925606 @default.
• W3217158093 creator A5076249543 @default.
• W3217158093 date "2021-12-01" @default.
• W3217158093 modified "2023-09-27" @default.
• W3217158093 title "Object-Aware Cropping for Self-Supervised Learning" @default.
• W3217158093 cites W1861492603 @default.
• W3217158093 cites W2010181071 @default.
• W3217158093 cites W2031489346 @default.
• W3217158093 cites W2108598243 @default.
• W3217158093 cites W2156163116 @default.
• W3217158093 cites W2163605009 @default.
• W3217158093 cites W2194775991 @default.
• W3217158093 cites W2577784528 @default.
• W3217158093 cites W2599837529 @default.
• W3217158093 cites W2765407302 @default.
• W3217158093 cites W2798991696 @default.
• W3217158093 cites W2804047946 @default.
• W3217158093 cites W2842511635 @default.
• W3217158093 cites W2926779152 @default.
• W3217158093 cites W2949517790 @default.
• W3217158093 cites W2951751045 @default.
• W3217158093 cites W2962742544 @default.
• W3217158093 cites W2962858109 @default.
• W3217158093 cites W2963420272 @default.
• W3217158093 cites W2963465221 @default.
• W3217158093 cites W2970597249 @default.
• W3217158093 cites W2992308087 @default.
• W3217158093 cites W3005680577 @default.
• W3217158093 cites W3018265077 @default.
• W3217158093 cites W3034781633 @default.
• W3217158093 cites W3035003500 @default.
• W3217158093 cites W3035164673 @default.
• W3217158093 cites W3035524453 @default.
• W3217158093 cites W3035682985 @default.
• W3217158093 cites W3036438747 @default.
• W3217158093 cites W3046208551 @default.
• W3217158093 cites W3047916742 @default.
• W3217158093 cites W3093274308 @default.
• W3217158093 cites W3093929102 @default.
• W3217158093 cites W3095121901 @default.
• W3217158093 cites W3102363610 @default.
• W3217158093 cites W3106428938 @default.
• W3217158093 cites W3107668149 @default.
• W3217158093 cites W3112137288 @default.
• W3217158093 cites W3118062200 @default.
• W3217158093 cites W3120167236 @default.
• W3217158093 cites W3121480429 @default.
• W3217158093 cites W3132401450 @default.
• W3217158093 cites W3135715136 @default.
• W3217158093 cites W3135958856 @default.
• W3217158093 cites W3149173402 @default.
• W3217158093 cites W3166784900 @default.
• W3217158093 cites W3168822201 @default.
• W3217158093 cites W3172615411 @default.
• W3217158093 cites W3185102306 @default.
• W3217158093 cites W639708223 @default.
• W3217158093 cites W7746136 @default.
• W3217158093 cites W3108262825 @default.
• W3217158093 hasPublicationYear "2021" @default.
• W3217158093 type Work @default.
• W3217158093 sameAs 3217158093 @default.
• W3217158093 citedByCount "0" @default.
• W3217158093 crossrefType "posted-content" @default.
• W3217158093 hasAuthorship W3217158093A5009078067 @default.
• W3217158093 hasAuthorship W3217158093A5015390901 @default.
• W3217158093 hasAuthorship W3217158093A5021971352 @default.
• W3217158093 hasAuthorship W3217158093A5023195442 @default.
• W3217158093 hasAuthorship W3217158093A5037733040 @default.
• W3217158093 hasAuthorship W3217158093A5059925606 @default.
• W3217158093 hasAuthorship W3217158093A5076249543 @default.
• W3217158093 hasConcept C118518473 @default.
• W3217158093 hasConcept C119857082 @default.
• W3217158093 hasConcept C13558536 @default.
• W3217158093 hasConcept C136389625 @default.
• W3217158093 hasConcept C153180895 @default.
• W3217158093 hasConcept C154945302 @default.
• W3217158093 hasConcept C166957645 @default.
• W3217158093 hasConcept C169258074 @default.
• W3217158093 hasConcept C17744445 @default.
• W3217158093 hasConcept C199360897 @default.
• W3217158093 hasConcept C199539241 @default.
• W3217158093 hasConcept C205649164 @default.
• W3217158093 hasConcept C2776151529 @default.
• W3217158093 hasConcept C2776359362 @default.
• W3217158093 hasConcept C2781238097 @default.
• W3217158093 hasConcept C41008148 @default.
• W3217158093 hasConcept C50644808 @default.
• W3217158093 hasConcept C75608658 @default.
• W3217158093 hasConcept C89600930 @default.
• W3217158093 hasConcept C94625758 @default.
• W3217158093 hasConceptScore W3217158093C118518473 @default.
• W3217158093 hasConceptScore W3217158093C119857082 @default.
• W3217158093 hasConceptScore W3217158093C13558536 @default.

• next