2022-05-10 更新
HierAttn: Effectively Learn Representations from Stage Attention and Branch Attention for Skin Lesions Diagnosis
Authors:Wei Dai, Rui Liu, Tianyi Wu, Min Wang, Jianqin Yin, Jun Liu
An accurate and unbiased examination of skin lesions is critical for the early diagnosis and treatment of skin cancers. The visual feature of the skin lesions varies significantly because skin images are collected from patients with different skin colours by using various devices. Recent studies have developed ensembled convolutional neural networks (CNNs) to classify the images for early diagnosis. However, the practical use of CNNs is limited because their network structures are heavyweight and neglect contextual information. Vision transformers (ViTs) learn the global features by self-attention mechanisms, but they also have comparatively large model sizes (more than 100M). To address these limitations, we introduce HierAttn, a lite and effective neural network with hierarchical and self attention. HierAttn applies a novel strategy based on learning local and global features by a multi-stage and hierarchical network. The efficacy of HierAttn was evaluated by using the dermoscopy images dataset ISIC2019 and smartphone photos dataset PAD-UFES-20. The experimental results show that HierAttn achieves the best top-1 accuracy and AUC among state-of-the-art mobile networks, including MobileNetV3 and MobileViT. The code is available at https://github.com/anthonyweidai/HierAttn.
PDF The code will be available at https://github.com/anthonyweidai/HierAttn as soon as possible
论文截图
ConvMAE: Masked Convolution Meets Masked Autoencoders
Authors:Peng Gao, Teli Ma, Hongsheng Li, Jifeng Dai, Yu Qiao
Vision Transformers (ViT) become widely-adopted architectures for various vision tasks. Masked auto-encoding for feature pretraining and multi-scale hybrid convolution-transformer architectures can further unleash the potentials of ViT, leading to state-of-the-art performances on image classification, detection and semantic segmentation. In this paper, our ConvMAE framework demonstrates that multi-scale hybrid convolution-transformer can learn more discriminative representations via the mask auto-encoding scheme. However, directly using the original masking strategy leads to the heavy computational cost and pretraining-finetuning discrepancy. To tackle the issue, we adopt the masked convolution to prevent information leakage in the convolution blocks. A simple block-wise masking strategy is proposed to ensure computational efficiency. We also propose to more directly supervise the multi-scale features of the encoder to boost multi-scale features. Based on our pretrained ConvMAE models, ConvMAE-Base improves ImageNet-1K finetuning accuracy by 1.4% compared with MAE-Base. On object detection, ConvMAE-Base finetuned for only 25 epochs surpasses MAE-Base fined-tuned for 100 epochs by 2.9% box AP and 2.2% mask AP respectively. Code and pretrained models are available at https://github.com/Alpha-VL/ConvMAE.
PDF 10 pages
论文截图