ALBERT

Description: This study examined the environmental impact of the three common organic pig husbandry systems, indoor (n = 24), partly outdoor (n = 30), and outdoor (n = 10), in eight European countries. Global warming (GWP), acidification (AP), and eutrophication potential (EP) was assessed per 1000 kg pig live weight on 64 farrow-to-finish pig production chains (cradle to farm gate). GWP, AP, and EP varied greatly, and the most important source was feed production, followed by housing. GWP did not differ between systems (p = 0.934), but AP in indoor systems and EP in outdoor systems were higher than in partly outdoor systems (p = 0.006 and p = 0.010, respectively). The higher AP in indoor systems can mainly be explained by NH3 arising from manure spreading, while PO4-eq arising from feed consumption and emissions on pasture accounted for the higher EP in outdoor systems. Associations of farm characteristics with (reduced) environmental impacts were mainly found for AP and EP, and included: (Increasing) farm size, numbers of piglets born and weaned per litter, (bought-in) mineral feed, and high-protein by-products, the latter probably connected to beneficial effects of appropriate dietary digestible lysine levels and feed conversion ratio. Increasing carcass weights and dietary cereal proportions were associated with higher environmental impacts. Overall, variation was mostly higher within than between systems, and measures to mitigate environmental impact were identified.

Description: High-quality data is necessary for modern machine learning. However, the acquisition of such data is difficult due to noisy and ambiguous annotations of humans. The aggregation of such annotations to determine the label of an image leads to a lower data quality. We propose a data-centric image classification benchmark with ten real-world datasets and multiple annotations per image to allow researchers to investigate and quantify the impact of such data quality issues. With the benchmark we can study the impact of annotation costs and (semi-)supervised methods on the data quality for image classification by applying a novel methodology to a range of different algorithms and diverse datasets. Our benchmark uses a two-phase approach via a data label improvement method in the first phase and a fixed evaluation model in the second phase. Thereby, we give a measure for the relation between the input labeling effort and the performance of (semi-)supervised algorithms to enable a deeper insight into how labels should be created for effective model training. Across thousands of experiments, we show that one annotation is not enough and that the inclusion of multiple annotations allows for a better approximation of the real underlying class distribution. We identify that hard labels can not capture the ambiguity of the data and this might lead to the common issue of overconfident models. Based on the presented datasets, benchmarked methods, and analysis, we create multiple research opportunities for the future directed at the improvement of label noise estimation approaches, data annotation schemes, realistic (semi-)supervised learning, or more reliable image collection.