期刊
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
卷 43, 期 2, 页码 567-578出版社
IEEE COMPUTER SOC
DOI: 10.1109/TPAMI.2019.2936841
关键词
Image coding; Training; Image reconstruction; Image analysis; Neural networks; Visualization; Task analysis; Gigapixel image analysis; computational pathology; convolutional neural networks; representation learning
资金
- Junior Researcher grant from the Radboud Institute of Health Sciences (RIHS), Nijmegen, The Netherlands
- Dutch Cancer Society [KUN 2015-7970]
- Dutch Cancer Society
- Alpe d'HuZes fund [KUN 2014-7032]
- European Union's Horizon 2020 research and innovation programme [825292]
The proposed Neural Image Compression (NIC) method utilizes weak image-level labels to build convolutional neural networks in two steps: compressing image representations and training CNN on these compressed representations. Through contrasting training and evaluation, it is found that NIC successfully integrates visual cues associated with image-level labels.
We propose Neural Image Compression (NIC), a two-step method to build convolutional neural networks for gigapixel image analysis solely using weak image-level labels. First, gigapixel images are compressed using a neural network trained in an unsupervised fashion, retaining high-level information while suppressing pixel-level noise. Second, a convolutional neural network (CNN) is trained on these compressed image representations to predict image-level labels, avoiding the need for fine-grained manual annotations. We compared several encoding strategies, namely reconstruction error minimization, contrastive training and adversarial feature learning, and evaluated NIC on a synthetic task and two public histopathology datasets. We found that NIC can exploit visual cues associated with image-level labels successfully, integrating both global and local visual information. Furthermore, we visualized the regions of the input gigapixel images where the CNN attended to, and confirmed that they overlapped with annotations from human experts.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据