Matches in SemOpenAlex for { <https://semopenalex.org/work/W3046208551> ?p ?o ?g. }
- W3046208551 abstract "Self-supervised representation learning approaches have recently surpassed their supervised learning counterparts on downstream tasks like object detection and image classification. Somewhat mysteriously the recent gains in performance come from training instance classification models, treating each image and it's augmented versions as samples of a single class. In this work, we first present quantitative experiments to demystify these gains. We demonstrate that approaches like MOCO and PIRL learn occlusion-invariant representations. However, they fail to capture viewpoint and category instance invariance which are crucial components for object recognition. Second, we demonstrate that these approaches obtain further gains from access to a clean object-centric training dataset like Imagenet. Finally, we propose an approach to leverage unstructured videos to learn representations that possess higher viewpoint invariance. Our results show that the learned representations outperform MOCOv2 trained on the same data in terms of invariances encoded and the performance on downstream image classification and semantic segmentation tasks." @default.
- W3046208551 created "2020-08-03" @default.
- W3046208551 creator A5053034244 @default.
- W3046208551 creator A5066334452 @default.
- W3046208551 date "2020-07-28" @default.
- W3046208551 modified "2023-09-27" @default.
- W3046208551 title "Demystifying Contrastive Self-Supervised Learning: Invariances, Augmentations and Dataset Biases" @default.
- W3046208551 cites W1536680647 @default.
- W3046208551 cites W1861492603 @default.
- W3046208551 cites W1903029394 @default.
- W3046208551 cites W1959608418 @default.
- W3046208551 cites W1991264156 @default.
- W3046208551 cites W2025768430 @default.
- W3046208551 cites W2054814877 @default.
- W3046208551 cites W2088049833 @default.
- W3046208551 cites W2105728138 @default.
- W3046208551 cites W2130325614 @default.
- W3046208551 cites W2145862222 @default.
- W3046208551 cites W2148349024 @default.
- W3046208551 cites W219040644 @default.
- W3046208551 cites W2250384498 @default.
- W3046208551 cites W2326925005 @default.
- W3046208551 cites W2467604901 @default.
- W3046208551 cites W2549418531 @default.
- W3046208551 cites W2575671312 @default.
- W3046208551 cites W2737258237 @default.
- W3046208551 cites W2794744029 @default.
- W3046208551 cites W2798991696 @default.
- W3046208551 cites W2842511635 @default.
- W3046208551 cites W2898200825 @default.
- W3046208551 cites W2944828972 @default.
- W3046208551 cites W2949517790 @default.
- W3046208551 cites W2962742544 @default.
- W3046208551 cites W2962824366 @default.
- W3046208551 cites W2962858109 @default.
- W3046208551 cites W2963420272 @default.
- W3046208551 cites W2963749571 @default.
- W3046208551 cites W2964700958 @default.
- W3046208551 cites W2972844547 @default.
- W3046208551 cites W2990873191 @default.
- W3046208551 cites W3005680577 @default.
- W3046208551 cites W3009561768 @default.
- W3046208551 cites W3034781633 @default.
- W3046208551 cites W3035524453 @default.
- W3046208551 cites W3106428938 @default.
- W3046208551 cites W3118062200 @default.
- W3046208551 cites W343636949 @default.
- W3046208551 hasPublicationYear "2020" @default.
- W3046208551 type Work @default.
- W3046208551 sameAs 3046208551 @default.
- W3046208551 citedByCount "33" @default.
- W3046208551 countsByYear W30462085512020 @default.
- W3046208551 countsByYear W30462085512021 @default.
- W3046208551 crossrefType "posted-content" @default.
- W3046208551 hasAuthorship W3046208551A5053034244 @default.
- W3046208551 hasAuthorship W3046208551A5066334452 @default.
- W3046208551 hasConcept C115961682 @default.
- W3046208551 hasConcept C119857082 @default.
- W3046208551 hasConcept C153083717 @default.
- W3046208551 hasConcept C153180895 @default.
- W3046208551 hasConcept C154945302 @default.
- W3046208551 hasConcept C17744445 @default.
- W3046208551 hasConcept C190470478 @default.
- W3046208551 hasConcept C199539241 @default.
- W3046208551 hasConcept C2776359362 @default.
- W3046208551 hasConcept C2781238097 @default.
- W3046208551 hasConcept C33923547 @default.
- W3046208551 hasConcept C37914503 @default.
- W3046208551 hasConcept C41008148 @default.
- W3046208551 hasConcept C64876066 @default.
- W3046208551 hasConcept C89600930 @default.
- W3046208551 hasConcept C94625758 @default.
- W3046208551 hasConceptScore W3046208551C115961682 @default.
- W3046208551 hasConceptScore W3046208551C119857082 @default.
- W3046208551 hasConceptScore W3046208551C153083717 @default.
- W3046208551 hasConceptScore W3046208551C153180895 @default.
- W3046208551 hasConceptScore W3046208551C154945302 @default.
- W3046208551 hasConceptScore W3046208551C17744445 @default.
- W3046208551 hasConceptScore W3046208551C190470478 @default.
- W3046208551 hasConceptScore W3046208551C199539241 @default.
- W3046208551 hasConceptScore W3046208551C2776359362 @default.
- W3046208551 hasConceptScore W3046208551C2781238097 @default.
- W3046208551 hasConceptScore W3046208551C33923547 @default.
- W3046208551 hasConceptScore W3046208551C37914503 @default.
- W3046208551 hasConceptScore W3046208551C41008148 @default.
- W3046208551 hasConceptScore W3046208551C64876066 @default.
- W3046208551 hasConceptScore W3046208551C89600930 @default.
- W3046208551 hasConceptScore W3046208551C94625758 @default.
- W3046208551 hasLocation W30462085511 @default.
- W3046208551 hasOpenAccess W3046208551 @default.
- W3046208551 hasPrimaryLocation W30462085511 @default.
- W3046208551 hasRelatedWork W1861492603 @default.
- W3046208551 hasRelatedWork W2108598243 @default.
- W3046208551 hasRelatedWork W2152790380 @default.
- W3046208551 hasRelatedWork W2194775991 @default.
- W3046208551 hasRelatedWork W2321533354 @default.
- W3046208551 hasRelatedWork W2326925005 @default.
- W3046208551 hasRelatedWork W2798991696 @default.
- W3046208551 hasRelatedWork W2842511635 @default.
- W3046208551 hasRelatedWork W2883725317 @default.