Conrad M Albrecht

Abstract:

Automatic label generation systems, which are capable to generate huge amounts of labels with limited human efforts, enjoy lots of potential in the deep learning era. These easy-to-come-by labels inevitably bear label noises due to a lack of human supervision and can bias model training to some inferior solutions. However, models can still learn some plausible features, before they start to overfit on noisy patterns. Inspired by this phenomenon, we propose a new Peaks fusion assisted EArly-Stopping (PEAS) approach for imagery segmentation with noisy labels, which is mainly composed of two parts. First, a fitting based early-stopping criterion is used to detect the turning phase from which models are about to mimic noise details. After that, a peaks fusion strategy is applied to select reliable models in the detection zone to generate final fusion results. Here, validation accuracies are utilized as indicators in model selection. The proposed method was evaluated on New York City dataset whose labels were automatically collected by a rule-based label generation system, thus noisy to some extent due to a lack of human supervision. The experimental results showed that the proposed PEAS method can achieve both promising statistical and visual results when trained with noisy labels.

Abstract:

Abstract. In this paper we benchmark a previously introduced big data platform that enables the analysis of big data from remote sensing and other geospatial-temporal data. The platform, called IBM PAIRS Geoscope, has been developed by leveraging open source big data technologies (Hadoop/HBase) that are in principle scalable in storage and compute to hundreds of PetaBytes. Currently, PAIRS hosts multiple PetaBytes of curated and geospatial-temporally indexed data. It organizes all data with key-value combinations, performing analytics close to the data to minimize data movement.