Matches in SemOpenAlex for { <https://semopenalex.org/work/W3196185920> ?p ?o ?g. }
- W3196185920 abstract "Automated animal censuses with aerial imagery are a vital ingredient towards wildlife conservation. Recent models are generally based on deep learning and thus require vast amounts of training data. Due to their scarcity and minuscule size, annotating animals in aerial imagery is a highly tedious process. In this project, we present a methodology to reduce the amount of required training data by resorting to self-supervised pretraining. In detail, we examine a combination of recent contrastive learning methodologies like Momentum Contrast (MoCo) and Cross-Level Instance-Group Discrimination (CLD) to condition our model on the aerial images without the requirement for labels. We show that a combination of MoCo, CLD, and geometric augmentations outperforms conventional models pre-trained on ImageNet by a large margin. Crucially, our method still yields favorable results even if we reduce the number of training animals to just 10%, at which point our best model scores double the recall of the baseline at similar precision. This effectively allows reducing the number of required annotations to a fraction while still being able to train high-accuracy models in such highly challenging settings." @default.
- W3196185920 created "2021-08-30" @default.
- W3196185920 creator A5005192117 @default.
- W3196185920 creator A5053177233 @default.
- W3196185920 creator A5055174808 @default.
- W3196185920 creator A5064396780 @default.
- W3196185920 creator A5069951807 @default.
- W3196185920 date "2021-08-17" @default.
- W3196185920 modified "2023-09-23" @default.
- W3196185920 title "Self-Supervised Pretraining and Controlled Augmentation Improve Rare Wildlife Recognition in UAV Images" @default.
- W3196185920 cites W1861492603 @default.
- W3196185920 cites W2108293993 @default.
- W3196185920 cites W2108598243 @default.
- W3196185920 cites W2138621090 @default.
- W3196185920 cites W2194775991 @default.
- W3196185920 cites W2321533354 @default.
- W3196185920 cites W2326925005 @default.
- W3196185920 cites W2342877626 @default.
- W3196185920 cites W2752508182 @default.
- W3196185920 cites W2798991696 @default.
- W3196185920 cites W2810030371 @default.
- W3196185920 cites W2842511635 @default.
- W3196185920 cites W2945020140 @default.
- W3196185920 cites W2953098847 @default.
- W3196185920 cites W2956586819 @default.
- W3196185920 cites W2962742544 @default.
- W3196185920 cites W2964246847 @default.
- W3196185920 cites W2991488782 @default.
- W3196185920 cites W3000080835 @default.
- W3196185920 cites W3009561768 @default.
- W3196185920 cites W3022061250 @default.
- W3196185920 cites W3034781633 @default.
- W3196185920 cites W3034978746 @default.
- W3196185920 cites W3035524453 @default.
- W3196185920 cites W3042772844 @default.
- W3196185920 cites W3095121901 @default.
- W3196185920 cites W3101821705 @default.
- W3196185920 cites W3105496885 @default.
- W3196185920 cites W3106485021 @default.
- W3196185920 cites W3108655343 @default.
- W3196185920 cites W3112898771 @default.
- W3196185920 cites W3120167236 @default.
- W3196185920 cites W3122325173 @default.
- W3196185920 cites W3123072794 @default.
- W3196185920 cites W3123939835 @default.
- W3196185920 cites W3136027703 @default.
- W3196185920 cites W3171007011 @default.
- W3196185920 cites W3175184835 @default.
- W3196185920 cites W3176455777 @default.
- W3196185920 cites W343636949 @default.
- W3196185920 cites W3097170885 @default.
- W3196185920 doi "https://doi.org/10.48550/arxiv.2108.07582" @default.
- W3196185920 hasPublicationYear "2021" @default.
- W3196185920 type Work @default.
- W3196185920 sameAs 3196185920 @default.
- W3196185920 citedByCount "0" @default.
- W3196185920 crossrefType "posted-content" @default.
- W3196185920 hasAuthorship W3196185920A5005192117 @default.
- W3196185920 hasAuthorship W3196185920A5053177233 @default.
- W3196185920 hasAuthorship W3196185920A5055174808 @default.
- W3196185920 hasAuthorship W3196185920A5064396780 @default.
- W3196185920 hasAuthorship W3196185920A5069951807 @default.
- W3196185920 hasBestOaLocation W31961859201 @default.
- W3196185920 hasConcept C100660578 @default.
- W3196185920 hasConcept C109747225 @default.
- W3196185920 hasConcept C111368507 @default.
- W3196185920 hasConcept C111919701 @default.
- W3196185920 hasConcept C119857082 @default.
- W3196185920 hasConcept C12725497 @default.
- W3196185920 hasConcept C127313418 @default.
- W3196185920 hasConcept C153180895 @default.
- W3196185920 hasConcept C154945302 @default.
- W3196185920 hasConcept C15744967 @default.
- W3196185920 hasConcept C162324750 @default.
- W3196185920 hasConcept C175444787 @default.
- W3196185920 hasConcept C180747234 @default.
- W3196185920 hasConcept C2776502983 @default.
- W3196185920 hasConcept C41008148 @default.
- W3196185920 hasConcept C774472 @default.
- W3196185920 hasConcept C98045186 @default.
- W3196185920 hasConceptScore W3196185920C100660578 @default.
- W3196185920 hasConceptScore W3196185920C109747225 @default.
- W3196185920 hasConceptScore W3196185920C111368507 @default.
- W3196185920 hasConceptScore W3196185920C111919701 @default.
- W3196185920 hasConceptScore W3196185920C119857082 @default.
- W3196185920 hasConceptScore W3196185920C12725497 @default.
- W3196185920 hasConceptScore W3196185920C127313418 @default.
- W3196185920 hasConceptScore W3196185920C153180895 @default.
- W3196185920 hasConceptScore W3196185920C154945302 @default.
- W3196185920 hasConceptScore W3196185920C15744967 @default.
- W3196185920 hasConceptScore W3196185920C162324750 @default.
- W3196185920 hasConceptScore W3196185920C175444787 @default.
- W3196185920 hasConceptScore W3196185920C180747234 @default.
- W3196185920 hasConceptScore W3196185920C2776502983 @default.
- W3196185920 hasConceptScore W3196185920C41008148 @default.
- W3196185920 hasConceptScore W3196185920C774472 @default.
- W3196185920 hasConceptScore W3196185920C98045186 @default.
- W3196185920 hasLocation W31961859201 @default.
- W3196185920 hasLocation W31961859202 @default.
- W3196185920 hasOpenAccess W3196185920 @default.