Improving Generalization of Deep Networks for Estimating Physical Properties of Containers and Fillings

TL;DR

This paper introduces a multimodal approach using a lightweight neural network with data augmentation and consistency measurement to improve the generalization of physical property estimation of household containers and fillings, especially for unseen objects.

Contribution

It proposes a novel combination of audio-video data, data augmentation, and consistency measurement to enhance model generalization for estimating container properties.

Findings

Improved accuracy in estimating container capacity, dimensions, and mass.

Enhanced generalization to unseen containers in the CCM dataset.

Effective filling type and level classification using combined audio-video data.

Abstract

We present methods to estimate the physical properties of household containers and their fillings manipulated by humans. We use a lightweight, pre-trained convolutional neural network with coordinate attention as a backbone model of the pipelines to accurately locate the object of interest and estimate the physical properties in the CORSMAL Containers Manipulation (CCM) dataset. We address the filling type classification with audio data and then combine this information from audio with video modalities to address the filling level classification. For the container capacity, dimension, and mass estimation, we present a data augmentation and consistency measurement to alleviate the over-fitting issue in the CCM dataset caused by the limited number of containers. We augment the training data using an object-of-interest-based re-scaling that increases the variety of physical values of the containers. We then perform the consistency measurement to choose a model with low prediction variance in the same containers under different scenes, which ensures the generalization ability of the model. Our method improves the generalization ability of the models to estimate the property of the containers that were not previously seen in the training.