Search

Automatic skin lesion analysis of dermoscopy images remains a challenging topic. In this paper, we propose an end-to-end multi-task deep learning framework for automatic skin lesion analysis. The proposed framework can perform skin lesion detection, classification, and segmentation tasks simultaneously. To address the class imbalance issue in the dataset (as often observed in medical image datasets) and meanwhile to improve the segmentation performance, a loss function based on the focal loss and the jaccard distance is proposed. During the framework training, we employ a three-phase joint training strategy to ensure the efficiency of feature learning. The proposed framework outperforms state-of-the-art methods on the benchmarks ISBI 2016 challenge dataset towards melanoma classification and ISIC 2017 challenge dataset towards melanoma segmentation, especially for the segmentation task. The proposed framework should be a promising computer-aided tool for melanoma diagnosis.

A growing number of state-of-the-art crowd counting methods employ the regression model. Such a model learns a person-density map first and its integral is further calculated to obtain the final count. However, this learned density map is uninterpretable and could deviate largely from the true person distribution even when the final count is accurate. In comparison, we present a conceptually interpretable and technically simple classification model for crowd counting, which consists of three novel modules: Deep Integrated Module (DIM), Scale Adaptive Module (SAM), and Interval Aware Module (IAM). Different from the traditional density map, the proposed pedestrian-aware density map (PADM) in our model can reveal the true people density, and meanwhile tackle the rarely-explored crowd localization task simultaneously. The proposed joint crowd counting and localization method does not require extra pretraining or fine-tuning for individual components of the network, and we train our model end-to-end in a single step. Without bells and whistles but a few lines of code, our simple yet effective method achieves better performances on both crowd counting and localization tasks when compared with state-of-the-art methods. The code is available online. [ABSTRACT FROM AUTHOR]

Previous transfer learning methods based on deep network assume the knowledge should be transferred between the same hidden layers of the source domain and the target domains. This assumption doesn't always hold true, especially when the data from the two domains are heterogeneous with different resolutions. In such case, the most suitable numbers of layers for the source domain data and the target domain data would differ. As a result, the high level knowledge from the source domain would be transferred to the wrong layer of target domain. Based on this observation, "where to transfer" proposed in this paper might be a novel research area. We propose a new mathematic model named DT-LET to solve this heterogeneous transfer learning problem. In order to select the best matching of layers to transfer knowledge, we define specific loss function to estimate the corresponding relationship between high-level features of data in the source domain and the target domain. To verify this proposed cross-layer model, experiments for two cross-domain recognition/classification tasks are conducted, and the achieved superior results demonstrate the necessity of layer correspondence searching. [ABSTRACT FROM AUTHOR]