Author: rsanthos

This week was a productive week for our team. We have continued training our model to improve our accuracy from about 70% to about 80%. We also made good progress in continuing to test and calibrate our ultrasonic sensors and connecting them to RPi. We also have started testing the compatibility of our iOS app with Apple’s accessibility features.

We ran into a risk this week. The Jetson Nano has suddenly started to be stuck on boot up and not proceed to its environment. Since the model has reached the end of life, there is very little help on this issue. We have temporarily switched to the Jetson Tx2 as there is more help for it, but we plan to try again with a different Jetson Nano concurrently. We prefer the Jetson Nano as its size works well for our product.

As a result, we are slightly behind schedule but hope to catch up this coming week. In addition, we haven’t made the decision to switch to the TX2 Jetson permanently, so our design remains the same.

Verifications and Validations
As a team, we hope to complete several validation tests this week. The first test we hope to do is on the latency. This end-to-end latency test will measure the time from when the Ultrasonic Sensor detects and object and when the audio message regarding the object is relayed to the user. We also hope the measure the time from when the camera takes a picture of an object and when the audio message on the object is relayed to the user. We hope to have a latency of 800 ms for both pipelines,

In addition, we hope to do user tests within the next two weeks. We hope to create a mock obstacle course and test the functionality of the product as users complete the obstacle course. We first hope to have the users do this obstacle course with no restrictions but solely for user feedback. With good success of this test, we hope to have users blindfolded and complete the obstacle course entirely relying on the product. The obstacle course will have several objects that we have trained our model for as well as objects that we have not. This will help us test objects that are known and objects that are unknown, but both should be detectable.

This week was a productive week for our team.  We have started training our model, made good progress in continuing to test and calibrate our ultrasonic sensors and connecting them to RPi and our Jetson, and also have started to work on our audio messages for our iOS app.

We ran into a small risk this week. While we were working on our audio messages, we realized that there might be a small compatibility issue for the text-to-speech for iOS 17. Switching to iOS 16 seems to have resolved the issue for the moment, but we will test extensively to ensure that this will not become an issue again.

The schedule has remained the same, and no design changes were made.

We haven’t run into any more risks as of this week.

One change we made was the program we are using to run the object detection ML model from YOLO v4 to YOLO v7-tiny. We have opted for this change in the model as the YOLO v7 model reduces computation and thus will reduce latency in the object detection model. Moreover,  the program works at a higher frame rate making it more accurate than the YOLO v4 model for object detection. Additionally, this model is more compatible with the RPi while maintaining a high accuracy. We haven’t incurred any costs as a result of this change, but we have benefited through lower latency and computation.

The schedule has remained the same.

 

A was written by Ryan, B was written by Oi and C was written by Ishan.

Part A:

When considering our product in a global context, our product hopes to bridge the gap in the ease of livelihood for people who are visually impaired compared to people who are not. Since 89% of visually impaired people live in low/middle income countries with over 62% in Asia, our product should significantly also help close the gap among the visually impaired community. With our goal to make our product affordable and function independently without the need for another human, we hope to help people in lower income countries travel easier, allowing them to accomplish more. In addition, as we develop our product we hope to help people travel to other countries as well (ie. navigating airport and flights) significantly increasing the opportunities for visually impaired people globally.

Source: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5820628/#:~:text=89%25%20of%20visually%20impaired%20people,East%20Asia%20(24%20million).

 

Part B:

There are many ways that our product solution is meeting specific needs with consideration of cultural factors. Many cultures place a high value on community support, inclusivity, and supporting those with disabilities. By helping the visually impaired navigate more independently, we are aligning with these values and fostering a more inclusive society. Next, there are some societies that have strong traditions of technological innovation and support for disability rights. Our product is a continuation of this tradition, where we use the latest technology to better social welfare. We will also be using the third most spoken language in the world, English, to provide voice over guidance to our users (https://www.babbel.com/en/magazine/the-10-most-spoken-languages-in-the-world). 

 

Part C:

When considering environmental factors, there are several different ways our product meets needs considering environmental factors. Our product can take into account extremities in the environment like fog or something that would make the camera quality of our device blurry by running our ML model on photos with different lighting and degrees of visibility.  Additionally, our device can enable visually impaired people to travel independently meaning that there’s less reliance on other modes of transport and other resources that could potentially damage the environment. Our device promotes and enables walking as a mode of transport, meaning less use of other modes of transport like cars that potentially damages the environment.