SIHeDA-Net: Sensor to Image Heterogeneous Domain Adaptation Network

Ishikaa Lunawat, Vignesh S, S P Sharan

Keywords: Domain Adaptation, Latent Space Transfer, American Sign Language

Abstract: The main advantage of wearable devices lies in enabling them to be tracked without external infrastructure. However, unlike vision (cameras), there is a dearth of large-scale training data to develop robust ML models for wearable devices. SIHeDA-Net (Sensor-Image Heterogeneous Domain Adaptation) uses training data from public images of American Sign Language (ASL) that can be used for inferences on sensors even with errors by bridging the domain gaps through latent space transfer.

Registration: I acknowledge that acceptance of this work at MIDL requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: novel methodological ideas without extensive validation

Primary Subject Area: Transfer Learning and Domain Adaptation

Secondary Subject Area: Learning with Noisy Labels and Limited Data

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.

TL;DR: Exploring heterogeneous domain transfer for gesture recognition. Our model, SIHeDA-Net, uses images of American Sign Language (ASL) as inferences on noisy sensor data by bridging the domain gaps through latent space transfer.

Code And Data: https://github.com/spider-tronix/SIHeDA-Net