CENSUS-HWR: a large training dataset for offline handwriting recognition
Authors:
Chetan Joshi,
Lawry Sorenson,
Ammon Wolfert,
Dr. Mark Clement,
Dr. Joseph Price,
Dr. Kasey Buckles
Abstract:
Progress in Automated Handwriting Recognition has been hampered by the lack of large training datasets. Nearly all research uses a set of small datasets that often cause models to overfit. We present CENSUS-HWR, a new dataset consisting of full English handwritten words in 1,812,014 gray scale images. A total of 1,865,134 handwritten texts from a vocabulary of 10,711 words in the English language…
▽ More
Progress in Automated Handwriting Recognition has been hampered by the lack of large training datasets. Nearly all research uses a set of small datasets that often cause models to overfit. We present CENSUS-HWR, a new dataset consisting of full English handwritten words in 1,812,014 gray scale images. A total of 1,865,134 handwritten texts from a vocabulary of 10,711 words in the English language are present in this collection. This dataset is intended to serve handwriting models as a benchmark for deep learning algorithms. This huge English handwriting recognition dataset has been extracted from the US 1930 and 1940 censuses taken by approximately 70,000 enumerators each year. The dataset and the trained model with their weights are freely available to download at https://censustree.org/data.html.
△ Less
Submitted 25 May, 2023;
originally announced May 2023.
Multi-criteria optimization and automated network restructuring to mitigate construction projects delays on-the-run
Authors:
Nina Prins,
Omar Kammouh,
A. R. M. Wolfert
Abstract:
Construction project management requires dynamic mitigation control ensuring the project's timely completion by a best fit for common purpose strategy for all stakeholders. Current mitigation approaches are usually performed by an iterative Monte Carlo (MC) analysis focussing on lowest-cost strategies which do not include (1) the project manager's goal-oriented behaviour, (2) automated network res…
▽ More
Construction project management requires dynamic mitigation control ensuring the project's timely completion by a best fit for common purpose strategy for all stakeholders. Current mitigation approaches are usually performed by an iterative Monte Carlo (MC) analysis focussing on lowest-cost strategies which do not include (1) the project manager's goal-oriented behaviour, (2) automated network restructuring potential, and (3) multi-dimensional optimization criteria for best fitting mitigation strategies-criteria. Therefore, the development statement within this paper is to design a method and implementation tool that properly dissolves all the aforementioned shortcomings ensuring the project's completion date by finding the most effective and efficient mitigation strategy. To fulfill the purpose of this paper, the Mitigation Controller (MitC) has been developed using an integrative approach of non-linear optimization techniques, probabilistic Monte Carlo simulation, and preference function modeling. Compared to the conventional way of mitigating project delays. The developed MitC allows mitigating potential delays with the least negative consequences on several project criteria, such as cost, environmental impact, etc. The application of the model to the demonstrative case study shows the ability of the model to significantly increase the probability of completing the project in the given target duration. Embedding the multi-criteria evaluation in the optimization model ensures that other interests are also represented in finding the optimal strategy for project delays.
△ Less
Submitted 20 June, 2022;
originally announced June 2022.
