MCIndoor20000: A fully-labeled image dataset to advance indoor objects detection

Abstract

A fully-labeled image dataset provides a unique resource for reproducible research inquiries and data analyses in several computational fields, such as computer vision, machine learning and deep learning machine intelligence. With the present contribution, a large-scale fully-labeled image dataset is provided, and made publicly and freely available to the research community. The current dataset entitled MCIndoor20000 includes more than 20,000 digital images from three different indoor object categories, including doors, stairs, and hospital signs. To make a comprehensive dataset addressing current challenges that exist in indoor objects modeling, we cover a multiple set of variations in images, such as rotation, intra-class variation plus various noise models. The current dataset is freely and publicly available at https://github.com/bircatmcri/MCIndoor20000.

A Highly Effective Deep Learning Based Escape Route Recognition Module for Autonomous Robots in Crisis and Emergency Situations

“…As demonstrated in table 2 our recognition module performs very well and achieves a very good classification performance. In addition, with an accuracy of 99.805 percent and a Kappa of 99.707 percent using the original dataset our module significantly outperforms the current benchmark by Bashiri et al [25] by 9.4 percent. In addition, our recognition module outperforms all existing classifiers in terms of classification accuracy (cf.…”

“…We aim to develop a recognition module with high performance which is suitable for embedding into operational robots and drones. The most important contributions are: 1) We build a highly effective recognition module with an accuracy of over 99.81 percent which significantly outperforms the current MCIndoor20000 benchmark [25]. 2) Our module architecture is 78 times smaller than the MCIndoor20000 benchmark model and needs a significantly smaller amount of computational power [25], making it suitable for embedding in artificial drones and robots [26].…”

“…The most important contributions are: 1) We build a highly effective recognition module with an accuracy of over 99.81 percent which significantly outperforms the current MCIndoor20000 benchmark [25]. 2) Our module architecture is 78 times smaller than the MCIndoor20000 benchmark model and needs a significantly smaller amount of computational power [25], making it suitable for embedding in artificial drones and robots [26]. 3) We extend the work by Dugdale's group on socially relevant multi-agent systems in crisis and emergency situations by giving the artificial agent the ability to precisely recognize objects [16].…”

“…Based on an own collected dataset from collapsed buildings they achieved accuracies from 76.17 to 83.64 percent. Bashiri et al [25] -who collected the MCIndoor20000 dataset which we also use in our work here -applied transfer learning (AlexNet) and achieved 90.40 percent accuracy on the original non-redundant part of their dataset. Giannakeris et al [30] presented a detection approach for classifying objects (e.g.…”

“…We evaluated our recognition module on the MCIn-door20000 dataset [25] consisting of 20,000 digital RGB images of hospital doors, signs and stairs. The dataset consists of two parts: the original dataset including 2,000 images and an extended dataset.…”