Matches in SemOpenAlex for { <https://semopenalex.org/work/W3208297503> ?p ?o ?g. }
- W3208297503 abstract "Recently contrastive learning has shown significant progress in learning visual representations from unlabeled data. The core idea is training the backbone to be invariant to different augmentations of an instance. While most methods only maximize the feature similarity between two augmented data, we further generate more challenging training samples and force the model to keep predicting discriminative representation on these hard samples. In this paper, we propose MixSiam, a mixture-based approach upon the traditional siamese network. On the one hand, we input two augmented images of an instance to the backbone and obtain the discriminative representation by performing an element-wise maximum of two features. On the other hand, we take the mixture of these augmented images as input, and expect the model prediction to be close to the discriminative representation. In this way, the model could access more variant data samples of an instance and keep predicting invariant discriminative representations for them. Thus the learned model is more robust compared to previous contrastive learning methods. Extensive experiments on large-scale datasets show that MixSiam steadily improves the baseline and achieves competitive results with state-of-the-art methods. Our code will be released soon." @default.
- W3208297503 created "2021-11-08" @default.
- W3208297503 creator A5004758964 @default.
- W3208297503 creator A5005672146 @default.
- W3208297503 creator A5045116626 @default.
- W3208297503 creator A5070487854 @default.
- W3208297503 date "2021-11-04" @default.
- W3208297503 modified "2023-10-16" @default.
- W3208297503 title "MixSiam: A Mixture-based Approach to Self-supervised Representation Learning" @default.
- W3208297503 cites W12634471 @default.
- W3208297503 cites W1520997877 @default.
- W3208297503 cites W2117539524 @default.
- W3208297503 cites W2138621090 @default.
- W3208297503 cites W2194775991 @default.
- W3208297503 cites W2321533354 @default.
- W3208297503 cites W2326925005 @default.
- W3208297503 cites W2599837529 @default.
- W3208297503 cites W2622263826 @default.
- W3208297503 cites W2757910899 @default.
- W3208297503 cites W2798991696 @default.
- W3208297503 cites W2842511635 @default.
- W3208297503 cites W2883725317 @default.
- W3208297503 cites W2891021639 @default.
- W3208297503 cites W2921861056 @default.
- W3208297503 cites W2944828972 @default.
- W3208297503 cites W2962742544 @default.
- W3208297503 cites W2962784289 @default.
- W3208297503 cites W2962824366 @default.
- W3208297503 cites W2963265008 @default.
- W3208297503 cites W2963399829 @default.
- W3208297503 cites W2963420272 @default.
- W3208297503 cites W2963465221 @default.
- W3208297503 cites W2963684088 @default.
- W3208297503 cites W2963826423 @default.
- W3208297503 cites W2970971581 @default.
- W3208297503 cites W2981381717 @default.
- W3208297503 cites W2987875759 @default.
- W3208297503 cites W2992308087 @default.
- W3208297503 cites W2998388430 @default.
- W3208297503 cites W3009561768 @default.
- W3208297503 cites W3012410440 @default.
- W3208297503 cites W3034781633 @default.
- W3208297503 cites W3034978746 @default.
- W3208297503 cites W3035058308 @default.
- W3208297503 cites W3035060554 @default.
- W3208297503 cites W3035224233 @default.
- W3208297503 cites W3035524453 @default.
- W3208297503 cites W3093423309 @default.
- W3208297503 cites W3095121901 @default.
- W3208297503 cites W3106428938 @default.
- W3208297503 cites W3106485021 @default.
- W3208297503 cites W3108655343 @default.
- W3208297503 cites W3114632476 @default.
- W3208297503 cites W3118608800 @default.
- W3208297503 cites W3128513979 @default.
- W3208297503 cites W3137863028 @default.
- W3208297503 cites W3171007011 @default.
- W3208297503 cites W3172615411 @default.
- W3208297503 cites W3203436460 @default.
- W3208297503 cites W343636949 @default.
- W3208297503 doi "https://doi.org/10.48550/arxiv.2111.02679" @default.
- W3208297503 hasPublicationYear "2021" @default.
- W3208297503 type Work @default.
- W3208297503 sameAs 3208297503 @default.
- W3208297503 citedByCount "1" @default.
- W3208297503 countsByYear W32082975032023 @default.
- W3208297503 crossrefType "posted-content" @default.
- W3208297503 hasAuthorship W3208297503A5004758964 @default.
- W3208297503 hasAuthorship W3208297503A5005672146 @default.
- W3208297503 hasAuthorship W3208297503A5045116626 @default.
- W3208297503 hasAuthorship W3208297503A5070487854 @default.
- W3208297503 hasBestOaLocation W32082975031 @default.
- W3208297503 hasConcept C103278499 @default.
- W3208297503 hasConcept C115961682 @default.
- W3208297503 hasConcept C119857082 @default.
- W3208297503 hasConcept C138885662 @default.
- W3208297503 hasConcept C153180895 @default.
- W3208297503 hasConcept C154945302 @default.
- W3208297503 hasConcept C177264268 @default.
- W3208297503 hasConcept C17744445 @default.
- W3208297503 hasConcept C190470478 @default.
- W3208297503 hasConcept C199360897 @default.
- W3208297503 hasConcept C199539241 @default.
- W3208297503 hasConcept C2776359362 @default.
- W3208297503 hasConcept C2776401178 @default.
- W3208297503 hasConcept C2776760102 @default.
- W3208297503 hasConcept C33923547 @default.
- W3208297503 hasConcept C37914503 @default.
- W3208297503 hasConcept C41008148 @default.
- W3208297503 hasConcept C41895202 @default.
- W3208297503 hasConcept C59404180 @default.
- W3208297503 hasConcept C94625758 @default.
- W3208297503 hasConcept C97931131 @default.
- W3208297503 hasConceptScore W3208297503C103278499 @default.
- W3208297503 hasConceptScore W3208297503C115961682 @default.
- W3208297503 hasConceptScore W3208297503C119857082 @default.
- W3208297503 hasConceptScore W3208297503C138885662 @default.
- W3208297503 hasConceptScore W3208297503C153180895 @default.
- W3208297503 hasConceptScore W3208297503C154945302 @default.
- W3208297503 hasConceptScore W3208297503C177264268 @default.