New turbid dataset from "Quantifying the degradation of optical algorithms in increasingly turbid mediums"

We’ve just posted two datasets from our Oceans 2021 paper “Quantifying the degradation of optical algorithms in increasingly turbid mediums”. Both datasets were taken in the UW-APL test tank. Data was recorded using ROS noetic.

Figure 1: Photo of Electrical Flying Lead (EFL) in UW-APL tank in low turbidity. Taken with custom Dalsa camera stereo unit.

Dataset #1 utilized our Numurus 3DX-C unit in a static configuration (i.e., without motion) as turbidity is induced using cornstarch. This dataset was utilized for the feature generation and correspondence analysis and depth map degradation analysis presented in our Oceans paper.

Dataset #2 was taken using a custom built stereo camera system using two Dalsa G3-GC11-C2420 Genie Nano machine vision cameras, as the cameras were swept through the scene. Turbidity was also increased over three turbid steps. This dataset was utilized for the deep learning analysis presented in our Oceans paper.

We encourage researchers to utilize this data to support research in optical algorithm degradation and mitigation in turbid environments.

Categories: dataset turbidity stereo numurus