Data Masking Analysis Based on Masked Autoencoders Architecture for Leaf Diseases Classification

Li, Lihua; Tanone, Radius

doi:10.1007/978-981-99-0605-5_7

Cited by 1 publication

(3 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We compare our proposed DropPos with the supervised pre-training baseline and a wide range of selfsupervised methods, including (i) contrastive learning methods [6,13], (ii) masked image modeling methods [3,28,34,54,56,64], and (iii) their combinations [19,72]. Effective pre-training epoch is used for fair comparison following [54,72] since it accounts for the actual trained images/views.…”

Section: Comparisons With Previous Resultsmentioning

confidence: 99%

“…This phenomenon can be mitigated by setting an appropriate τ . A large τ leads [CVPR '22] 1600 83.6 85.9 SimMIM [64] [CVPR '22] 800 83.8 -SemMAE [34] [NeurIPS '22] 800 83.4 -LocalMIM [56] [CVPR '23] 1600 84.0 -HPM [54] [CVPR '23] 800 84.2 85.8…”

Section: Ablation Studiesmentioning

confidence: 99%

“…When fine-tuning using backbones pre-trained with different methods, we search for the optimal learning rate or simply follow their official implementation for a fair comparison. Specifically, the learning rate is 1e-4 for [13,28,34], 4e-4 for [19,56], respectively. All experiments are conducted on 8 Tesla V100 GPUs.…”

Section: A3 Ade20k Semantic Segmentationmentioning

confidence: 99%

See 2 more Smart Citations

Key Technology of Automation Control Based on Artificial Intelligence Technology

Wang

2021

Advances in Intelligent Systems and Computing

View full text Add to dashboard Cite

Current state-of-the-art approaches for few-shot action recognition achieve promising performance by conducting frame-level matching on learned visual features. However, they generally suffer from two limitations: i) the matching procedure between local frames tends to be inaccurate due to the lack of guidance to force long-range temporal perception; ii) explicit motion learning is usually ignored, leading to partial information loss. To address these issues, we develop a Motion-augmented Long-short Contrastive Learning (MoLo) method that contains two crucial components, including a long-short contrastive objective and a motion autodecoder. Specifically, the long-short contrastive objective is to endow local frame features with long-form temporal awareness by maximizing their agreement with the global token of videos belonging to the same class. The motion autodecoder is a lightweight architecture to reconstruct pixel motions from the differential features, which explicitly embeds the network with motion dynamics. By this means, MoLo can simultaneously learn long-range temporal context and motion cues for comprehensive few-shot matching. To demonstrate the effectiveness, we evaluate MoLo on five standard benchmarks, and the results show that MoLo favorably outperforms recent advanced methods. The source code is available at https://github. com/alibaba-mmai-research/MoLo.

show abstract