Assisting guide dogs with data: Guiding Eyes for the Blind
A team from North Carolina State University and IBM are working with the non-profit to improve operations
Guiding Eyes for the Blind is one of the leading non-profits for providing guide dogs to those with vision loss. Guiding Eyes has the goal to raise, train, and match as many dogs as possible. A successful pairing means many years of life changing connections giving the confidence to those with vision loss to live more independently. This is no easy feat as a good match requires a dog that has personality traits matched to their handler as well as the ability to stay calm under stressful conditions, like crossing a busy street or walking through a park filled with distractions..
Guiding Eyes has been partnering with IBM and North Carolina State University (NCSU) to use technology and data on a journey to improve its puppy development and breeding programs to increase the success for its dogs. By using hardware, software, and data analysis to explore areas from mapping the personality traits of 7-week-old puppies they are monitoring how dogs perform under stress. New-meets-old in this case as an organization that has been around since the 1950s is using cloud native technologies to improve its operations. You can get a sense for how much data is generated as a dog goes through milestones in its development in this video:
Understanding a dog’s vitals: harnesses and collars
Alper Bozkurt, a Professor at NCSU, is an expert in sensors and biological instrumentation systems, and with his team has been developing tools to gather vital information from dogs. Whereas typically research projects will use a sample size of ten or so, his team has been pushing the boundaries by launching a project that will deploy hundreds of collars to volunteers raising potential guide dogs, environmental factors, and location. These measurements provide valuable data on key insights. These could include how active a dog is on a daily basis, the influence of frequency and pattern of socialization on success and understanding whether a dog gets more agitated in certain situations.
Professor Bozkurt’s team has also developed a more robust hardware harness that can be used in the more controlled environment of the Guiding Eyes training facility. This device requires some more setup than the typical harness, but provides deeper information around physiology and behavior including heart rate, respiratory rate, and movement.
As a note for those concerned: the team went through great lengths to ensure the puppy participants’ welfare was the top priority. All puppies can “opt out” if they show any stress wearing the harness according to a protocol developed in consultation with a veterinary behaviorist.
Turning dog sensor data into insights
All of the sensor data coming in from dog collars and harnesses needs to be understood and turned into insights that can influence the decisions that the Guiding Eyes organization makes. Dave Roberts is an Associate Professor at NCSU and CEO of Organic Intelligence Technologies, with strong expertise in behavior analysis, whether that’s on humans or other species.
Working closely with Professor Bozkurt and the IBM Center for Advanced Studies (CAS) team led by Andy Rindos, Professor Roberts has developed an artificial intelligence platform built on a Kubernetes platform and other cloud-based assets, all running within the IBM Cloud for Education, to analyze the large volume of data coming from the collars and harnesses. Using the IBM Watson Studio Machine Learning service for data ingress and analysis, his team has been able to start producing insights for Guiding Eyes. The configuration of these assets has been preserved within the IBM Cloud for Education for reuse by academicians interested in performing similar research.
Dog temperament for a puppy has been identified by Guiding Eyes as an important factor to determine whether a dog makes it successfully through the training program and goes on to help a person in need. Resources are focused on dogs that really enjoy guide dog work and others find other roles, such as being placed in public adoption programs.
Typically, a Guiding Eyes trainer goes through five years of training to look at a litter of puppies and determine the temperament using a detailed checklist. Early findings from the team’s AI platform have been encouraging in this regard. By using the data input from the harnesses, Professor Roberts and the CAS team have been able to predict with 94% accuracy the temperament of a dog compared to the assessment by an expert trainer. These results open up the potential for systems like this to help guide dog training organizations around the world supplement the role of trainers. The team is continuing to build on these early findings and looking for other ways to measure behavior and other determining factors that can help non-profits like Guiding Eyes improve their operations.
Bringing AI into areas like guide dog training and health monitoring opens up new possibilities for improving our relationship with exceptional animals. As Lorraine Trapani, an IBM volunteer with Guiding Eyes and advocate for this project points out,
“I’m excited to be involved with the collection of biometric data from the puppies for the application of machine learning and analytics — as ultimately this will help more puppies — and volunteer raisers — succeed. It’s an exciting opportunity to use AI, first hand, in a way that benefits society.”