Wireless Potato Root Tuber Sensing 1.0 (WPS 1.0)

We introduce a wireless potato root tuber sensing (WPS) dataset that comprises multi-channel received signal strength (RSS) data collected from a wireless network and ground truth annotations of potato tubers. We design a testbed named Spin, which is implemented using a multi-channel wireless network. Specifically, the wireless network consists of 16 TI CC2531 sensor nodes deployed around the sensing area, each of which operates on the 2.4GHz frequency band with 16 distinct frequency channels.  Using the testbed, we perform extensive measurement experiments. First, we perform experiments in a static environment (Case 1) to collect RSS data from underground potato tubers placed at different positions. Second, considering the impact of dynamic environmental changes on wireless signals, we perform two sets of experiments to investigate the effects of human activities and changes in environmental layout (Case 2). Third, we perform cross-environment and cross-soil experiments (Case 3) to investigate the interference from different environments and different soil moisture levels for wireless signals. The experimental environments include a hallway, a meeting room, and a living room, while the soil moisture levels are 7.1% and 11.2%. Using the collected RSS data and ground truth annotations, we develop a deep learning model for reconstructing the maximum cross-section images of underground potato root tubers. The source code and pre-trained models are publicly available at https://github.com/Data-driven-RTI/MC_Diffusion.

The dataset's file structure is depicted as follows:

The data in the Case 1 folder are collected in a static environment and include ten differently sized potatoes, each of which is rotated 28 times across three randomly selected positions. Thus, for Case 1, the file naming convention is as follows: tuberID_positionID_rotationID. The data in the Case 2 folder are collected under dynamic environmental conditions and include two types of changes:  environmental layout alterations and human activity. The "EnvironmentChange" folder contains RSS data and corresponding ground truth obtained during environmental layout changes. The environmental layouts before and after the change are labeled as E1 and E2, respectively.  Each folder contains RSS data from 26 potato tubers, each placed at a fixed position within the sensing area. The file naming convention is as follows: tuberID. The "HumanActivity" folder contains RSS data and corresponding ground truth under human activity conditions. We select four positions to place six tubers sequentially, and a person performs an activity around the sensing area at four different distances during data collection. Thus, the file naming convention is as follows: positionID_tuberID_distanceID. The data in the Case 3 folder are collected across different environments and different soil moisture levels. The tubers are placed at different positions, and the file naming follows positionID_tuberID. The "Code.zip" file contains data processing scripts that segment the original data into individual samples, each comprising RSS values collected from 16 sensor nodes across 16 frequency channels.