Self-supervised Transparent Liquid Segmentation for Robotic Pouring

Robots equipped to pour liquids would support in responsibilities like cooking or watering our plants. Having said that, clear liquids are difficult to understand in images. A modern paper released on proposes a strategy for perceiving clear liquid within clear containers.

Transparent liquid in a transparent container.

Transparent liquid in a clear container. Image credit rating: Piqsels, CC0 Community Domain

The approach employs a generative product that learns to translate visuals of coloured liquid into artificial illustrations or photos of a clear liquid, which can be utilised to coach a transparent liquid segmentation product. It does not have to have labeled correspondence concerning colored photos and clear visuals and as a result permits automatic and hugely successful dataset collection.

Scientists assemble a robotic pouring process to show the utility of the clear liquid segmentation design. Furthermore, a number of dataset augmentation experiments are carried out to demonstrate the possible of the proposed method to generalize to assorted scenes.

Liquid state estimation is crucial for robotics responsibilities such as pouring however, estimating the condition of clear liquids is a difficult problem. We suggest a novel segmentation pipeline that can section transparent liquids this sort of as water from a static, RGB impression with no requiring any manual annotations or heating of the liquid for teaching. As an alternative, we use a generative product that is able of translating illustrations or photos of colored liquids into synthetically created clear liquid pictures, properly trained only on an unpaired dataset of colored and transparent liquid photographs. Segmentation labels of coloured liquids are attained quickly working with history subtraction. Our experiments demonstrate that we are able to precisely predict a segmentation mask for clear liquids without the need of requiring any guide annotations. We demonstrate the utility of transparent liquid segmentation in a robotic pouring activity that controls pouring by perceiving the liquid peak in a transparent cup.

Exploration paper: Narayan Narasimhan, G., Zhang, K., Eisner, B., Lin, X., and Held, D., “Self-supervised Clear Liquid Segmentation for Robotic Pouring”, 2022. Connection: muscles/2203.01538

Rosa G. Rose

Next Post

Fusion Plasmas Meet Their Match in Reinforcement Learning

Tue Mar 8 , 2022
Identical mismatches in source and demand from customers contributed to significant cascading blackouts in August 2003 in the northeastern United States and Canada, in July 2012 in India, and in March 2019 in Venezuela. The condition is not likely to get better whenever soon, for a few causes. To start […]