Matches in SemOpenAlex for { <https://semopenalex.org/work/W3217621203> ?p ?o ?g. }
- W3217621203 abstract "Videos are a rich source for self-supervised learning (SSL) of visual representations due to the presence of natural temporal transformations of objects. However, current methods typically randomly sample video clips for learning, which results in an imperfect supervisory signal. In this work, we propose PreViTS, an SSL framework that utilizes an unsupervised tracking signal for selecting clips containing the same object, which helps better utilize temporal transformations of objects. PreViTS further uses the tracking signal to spatially constrain the frame regions to learn from and trains the model to locate meaningful objects by providing supervision on Grad-CAM attention maps. To evaluate our approach, we train a momentum contrastive (MoCo) encoder on VGG-Sound and Kinetics-400 datasets with PreViTS. Training with PreViTS outperforms representations learnt by contrastive strategy alone on video downstream tasks, obtaining state-of-the-art performance on action classification. PreViTS helps learn feature representations that are more robust to changes in background and context, as seen by experiments on datasets with background changes. Learning from large-scale videos with PreViTS could lead to more accurate and robust visual feature representations." @default.
- W3217621203 created "2021-12-06" @default.
- W3217621203 creator A5017855052 @default.
- W3217621203 creator A5018518655 @default.
- W3217621203 creator A5046238088 @default.
- W3217621203 creator A5060883710 @default.
- W3217621203 creator A5081479937 @default.
- W3217621203 date "2021-12-01" @default.
- W3217621203 modified "2023-10-14" @default.
- W3217621203 title "PreViTS: Contrastive Pretraining with Video Tracking Supervision" @default.
- W3217621203 cites W1520997877 @default.
- W3217621203 cites W1836533770 @default.
- W3217621203 cites W1861492603 @default.
- W3217621203 cites W2031489346 @default.
- W3217621203 cites W2034014085 @default.
- W3217621203 cites W2108598243 @default.
- W3217621203 cites W2117539524 @default.
- W3217621203 cites W2126579184 @default.
- W3217621203 cites W2138621090 @default.
- W3217621203 cites W2148349024 @default.
- W3217621203 cites W219040644 @default.
- W3217621203 cites W2198618282 @default.
- W3217621203 cites W24089286 @default.
- W3217621203 cites W2470139095 @default.
- W3217621203 cites W2487442924 @default.
- W3217621203 cites W2511428026 @default.
- W3217621203 cites W2575671312 @default.
- W3217621203 cites W2769833683 @default.
- W3217621203 cites W2798991696 @default.
- W3217621203 cites W2799087757 @default.
- W3217621203 cites W2842511635 @default.
- W3217621203 cites W2883451034 @default.
- W3217621203 cites W2887997457 @default.
- W3217621203 cites W2895243423 @default.
- W3217621203 cites W2902016181 @default.
- W3217621203 cites W2948242301 @default.
- W3217621203 cites W2962858109 @default.
- W3217621203 cites W2962931121 @default.
- W3217621203 cites W2963524571 @default.
- W3217621203 cites W2963749571 @default.
- W3217621203 cites W2964037671 @default.
- W3217621203 cites W2970642899 @default.
- W3217621203 cites W2971155163 @default.
- W3217621203 cites W2979579363 @default.
- W3217621203 cites W2991133498 @default.
- W3217621203 cites W3009561768 @default.
- W3217621203 cites W3012540512 @default.
- W3217621203 cites W3015371781 @default.
- W3217621203 cites W3034781633 @default.
- W3217621203 cites W3034978746 @default.
- W3217621203 cites W3035058308 @default.
- W3217621203 cites W3035524453 @default.
- W3217621203 cites W3094454579 @default.
- W3217621203 cites W3095121901 @default.
- W3217621203 cites W3100859887 @default.
- W3217621203 cites W3105422445 @default.
- W3217621203 cites W3108655343 @default.
- W3217621203 cites W3128637142 @default.
- W3217621203 cites W3135958856 @default.
- W3217621203 cites W3145659481 @default.
- W3217621203 cites W3168796319 @default.
- W3217621203 cites W3174308973 @default.
- W3217621203 cites W3176166976 @default.
- W3217621203 cites W3181598125 @default.
- W3217621203 cites W3182683290 @default.
- W3217621203 cites W3187567282 @default.
- W3217621203 cites W3204261852 @default.
- W3217621203 cites W3204659849 @default.
- W3217621203 doi "https://doi.org/10.48550/arxiv.2112.00804" @default.
- W3217621203 hasPublicationYear "2021" @default.
- W3217621203 type Work @default.
- W3217621203 sameAs 3217621203 @default.
- W3217621203 citedByCount "0" @default.
- W3217621203 crossrefType "posted-content" @default.
- W3217621203 hasAuthorship W3217621203A5017855052 @default.
- W3217621203 hasAuthorship W3217621203A5018518655 @default.
- W3217621203 hasAuthorship W3217621203A5046238088 @default.
- W3217621203 hasAuthorship W3217621203A5060883710 @default.
- W3217621203 hasAuthorship W3217621203A5081479937 @default.
- W3217621203 hasBestOaLocation W32176212031 @default.
- W3217621203 hasConcept C111919701 @default.
- W3217621203 hasConcept C118505674 @default.
- W3217621203 hasConcept C138885662 @default.
- W3217621203 hasConcept C151730666 @default.
- W3217621203 hasConcept C153180895 @default.
- W3217621203 hasConcept C154945302 @default.
- W3217621203 hasConcept C199360897 @default.
- W3217621203 hasConcept C202474056 @default.
- W3217621203 hasConcept C2776401178 @default.
- W3217621203 hasConcept C2779343474 @default.
- W3217621203 hasConcept C2779843651 @default.
- W3217621203 hasConcept C2781238097 @default.
- W3217621203 hasConcept C28490314 @default.
- W3217621203 hasConcept C31972630 @default.
- W3217621203 hasConcept C41008148 @default.
- W3217621203 hasConcept C41895202 @default.
- W3217621203 hasConcept C59404180 @default.
- W3217621203 hasConcept C86803240 @default.
- W3217621203 hasConceptScore W3217621203C111919701 @default.
- W3217621203 hasConceptScore W3217621203C118505674 @default.