In digital histopathology, entire neoplasm segmentation on Whole Slide Image (WSI) of Hepatocellular Carcinoma (HCC) plays an important role, especially as a preprocessing filter to automatically exclude healthy tissue, in histological molecular correlations mining and other downstream histopathological tasks. The segmentation task remains challenging due to HCC's inherent high-heterogeneity and the lack of dependency learning in large field of view. In this article, we propose a novel deep learning architecture with a hierarchical Transformer encoder, HiTrans, to learn the global dependencies within expanded 4096$\times$4096 WSI patches. HiTrans is designed to encode and decode the patches with larger reception fields and the learned global dependencies, compared to the state-of-the-art Fully Convolutional Neural networks (FCNN). Empirical evaluations verified that HiTrans leads to better segmentation performance by taking into account regional and global dependency information.
Temperature is a widely used hyperparameter in various tasks involving neural networks, such as classification or metric learning, whose choice can have a direct impact on the model performance. Most of existing works select its value using hyperparameter optimization methods requiring several runs to find the optimal value. We propose to analyze the impact of temperature on classification tasks by describing a dataset as a set of statistics computed on representations on which we can build a heuristic giving us a default value of temperature. We study the correlation between these extracted statistics and the observed optimal temperatures. This preliminary study on more than a hundred combinations of different datasets and features extractors highlights promising results towards the construction of a general heuristic for temperature.
The aim of this study is to detect man-made cartographic objects in high-resolution satellite images. New generation satellites offer a sub-metric spatial resolution, in which it is possible (and necessary) to develop methods at object level rather than at pixel level, and to exploit structural features of objects. With this aim, a method to generate structural object models from manually segmented images has been developed. To generate the model from non-segmented images, extraction of the objects from the sample images is required. A hybrid method of extraction (both in terms of input sources and segmentation algorithms) is proposed: A region based segmentation is applied on a 10 meter resolution multi-spectral image. The result is used as marker in a "marker-controlled watershed method using edges" on a 2.5 meter resolution panchromatic image. Very promising results have been obtained even on images where the limits of the target objects are not apparent.