This experiment is part of the AI for Interspecies Communication Challenge Grant. Browse more projects

Sensing canine behavior from an edge ML-driven haptic interface

Raised of $4,520 Goal
Funded on 11/20/23
Successfully Funded
  • $4,610
  • 101%
  • Funded
    on 11/20/23

About This Project

Edge machine learning refers to the process of running embedded ML models on site using devices capable of collecting, processing, and recognizing patterns within collections of raw data. This project seeks to train one of such devices (Nicla SenseME) with dog data from the Earth Species' Bio-logger Ethogram Benchmark (BEBE). The board will be used as a smart dog collar, with its ML inferences controlling haptics vibrations in a bracelet wore by a person.

Ask the Scientists

Join The Discussion

What is the context of this research?

"Growing symbiotic new senses for humans" sets the context here. In experimenting with haptic interfaces to grant humans a direct perceptual experience of a dog's activities, we started to consider how edge ML could help to identify patterns in accelerometer data from a dog's tracker.

We've been able to create a prototype that demonstrates our vision for the use of edge sensing and embedded ML processing. However, we have used an open source ML model that was trained on human data. So the next step is to design an edge ML model with actual dog data. More specifically, data from "Movement Sensor Dataset for Dog Behavior Classification", Vehkaoja et al (2021), as described by the Bio-logger Ethogram Benchmark.

What is the significance of this project?

In our short paper "A Human-Canine Interface From Sensors, Haptics, and AI", we have outlined our larger vision and its main applications. Namely, the possibilities of such an interface to expand our direct perception of reality in relation to dogs, which could have immediate implications for:

1. Blind people who rely on guide dogs for locomotion.

2. Professionals who rely on dog's olfactory abilities to detect objects and even diseases (yes, dogs can detect 28+ diseases in real-time, often with higher accuracy than our best diagnostics).

Furthermore, Edge ML represents a great promise to non-human understanding from sensory data, especially due to on-site data processing. From conservation to sensory addition, this has been a key factor and our work is well placed to explore this frontier.

What are the goals of the project?

Our project begins by accessing and preprocessing canine data from "Movement Sensor Dataset for Dog Behavior Classification" by Vehkaoja et al (2021). Using Edge Impulse Studio, we'll train a machine learning model, ensuring data aligns with the "Data acquisition formatspecification". We'll test our model with a smart tracker on Maniçoba, specifically focusing on data from the neck and torso, as done in the original paper.

A pivotal aspect of our exploration is to discern the long-term effects of vibro-tactile feedback derived from canine behavior on human sensory perception. This could be especially relevant in the context of enhanced experiences for the visually impaired.

Our journey and findings will be summarized in a paper, detailing our workflow from data collection to haptic feedback.


Please wait...

Considering that this project will be conducted in the same independent fashion as its predecessor, the same logic was used for budget estimates:

Researcher Expenses are allocated for the researchers' dedicated time and effort for approximately 3 months. This includes tasks such as hardware assembling, data analysis, model training, tuning, and comprehensive project documentation. Lastly, Platform Fees account for the necessary costs of using this fundraising platform, including an 8% platform fee and a 3-5% payment processing fee.

Endorsed by

I have engaged with Danilo in the past year, as he has been using the haptics platform developed by our team at Google. This project is an insightful exploration that connects biology, art, AI, and engineering. Such a project is timely now as many natural resources become scarce and endangered; we need AI-enabled tools to better understand and appreciate the natural world around us. Danilo is the right person to take on this difficult challenge with his right mix of multidisciplinary background, curiosity, and motivation.

Project Timeline

R&D is already in motion with literature review, equipment assembly and preliminary data processing. The funding campaign should run until mid October, and once funding is secured and received I can proceed with deliverables as detailed in the timeline:

Sep 22, 2023

Project Launched

Oct 21, 2023

Project Launched

Oct 31, 2023

Secure funds

Nov 08, 2023

Do any necessary data preprocessing 

Nov 15, 2023

Train and test first ML models using Edge Impulse studio

Meet the Team

Independent Researcher

View Profile
Akash Kulgod
Akash Kulgod

Team Bio

We are a guild of interdependent artists and researchers working towards S2S (Species-to-Species) InterComVerse, through the development of novel technological products, as well as immersive experiences.

  • Inter... connection, being, relation
  • Com... munication, munion, munity
  • Verse... poetry making
  • Danilo

    Weaver of people, ideas, and creative flows. Life transformed by extraordinary educational opportunities (scholarships for: high school at Colégio Sidarta, and undergrad program in a "#1 most innovative university in the world"). Since 2015, digital nomad. Since 2017, freelancer and independent researcher of the Complexity Sciences. Since 2020, father, neo-rural and instructor in digital education platforms. Recently, self-proclaimed amateur artist.

    Project Backers

    • 5Backers
    • 101%Funded
    • $4,610Total Donations
    • $922.00Average Donation
    Please wait...