The state of the art in semantic segmentation is steadily increasing in performance, resulting in more precise and reliable segmentations in many different applications. However, progress is limited by the cost of generating labels for training, which sometimes requires hours of manual labor for a single image. Because of this, semi-supervised methods have been applied to this task, with varying degrees of success. A key challenge is that common augmentations used in semi-supervised classification are less effective for semantic segmentation. We propose a novel data augmentation mechanism called ClassMix, which generates augmentations by mixing unlabelled samples, by leveraging on the network's predictions for respecting object boundaries. We evaluate this augmentation technique on two common semi-supervised semantic segmentation benchmarks, showing that it attains state-of-theart results. Lastly, we also provide extensive ablation studies comparing different design decisions and training regimes. Code is available at: https://github.com/WilhelmT/ClassMix.

Semantic segmentation models based on convolutional neural networks have recently displayed remarkable performance for a multitude of applications. However, these models typically do not generalize well when applied on new domains, especially when going from synthetic to real data. Unsupervised domain adaptation (UDA) attempts to train on labelled data from one domain (source domain), and simultaneously learn from unlabelled data in the domain of interest (target domain). Existing methods have seen success by training on pseudo-labels for these unlabelled images. Multiple techniques have been proposed to mitigate low-quality pseudo-labels arising from the domain shift, with varying degrees of success. We propose DACS: Domain Adaptation via Cross-domain mixed Sampling, which mixes images from the two domains along with the corresponding labels. These mixed samples are then trained on, in addition to the labelled data itself. We demonstrate the effectiveness of our solution by achieving state-of-the-art results for two common synthetic-toreal semantic segmentation benchmarks for UDA. Code is available at: https://github.com/ vikolss/DACS.

The human interpretation of analytical outputs is a significant challenge in forensic science, making it vital to explore the application of protocols as we enhance our practices. This study assesses decision making in forensic anthropological analyses utilizing eye-tracking technology to quantify an observer’s estimate of confidence and reliability. Ten individuals with varying levels of education and experience were asked to score cranial morphologies for two human crania. Each participant’s fixation points, fixation duration, and visit count and duration were assessed using Tobii™ Pro 2 eye-tracking glasses. Mid-facial morphologies capturing relative widths were the quickest scored traits, with an overall median time of 14.59 seconds; more complex morphological assessments took longer. Using time as a proxy for confidence, Kruskal-Wallis rank sum results indicate individuals with less experience differed significantly from individuals with greater experience (p = 0.01) although differences in level of education were not significant. Interestingly, intraclass correlation coefficients (ICC) indicate interobserver reliability is high between observers, suggesting experience only slightly improves agreement. These preliminary results suggest experience is more important than level of education. Through empirical decision making studies, forensic anthropologists can improve practices—increasing the transparency of evaluative decision making by targeting confusing or problematic aspects of a data collection practice, and in so doing, enhance training.