Nii-Okaitey Tetteh, Chi Nguyen, Likhitha Chintareddy | Carnegie Mellon ECE Capstone, Spring 2023
I’ve finished fixing the errors on the CV side and fine-tuning it so the code should be ready to run on the Jetson Nano. Testing the code is not a one-time task and it’s something that I will keep doing every week. As we’ve finally …
This week I received the rest of the parts I ordered before break. Now that I have all the hardware components, I began setting up the Jetson. I got the OS working on the Jetson, and cloned our github repo onto the board. Additionally, I …
In the earlier week, I went through the results of the user studies conducted with various design ideas and came across an interesting paper [linked HERE] that led to a massive redesign of our initial idea. This resulted in our idea going from a simple line-based representation of the number of people waiting to a more visually informative design inspired by the study
Further information on how and why this design is interesting and fits our users’ needs was given in the Design Report. Later in the week, I laid down some preliminary work to rewrite the code for the UI (as our design has been changed). However, due to the nature of spring break and a flu I caught in the earlier part of the week, work has been delayed. I’m currently still working on the UI but will have to simultaneously start working on the AJAX implementation in the later part of the week. This gives me about 4-5 days to finish major sections of the UI where I can focus solely on it. So I’m delayed from the schedule by 4-5 days but working on the AJAX in the later part of the week including the weekends should hopefully make sure I will still be on track without necessarily having to push anything in the schedule.
To answer the question for report #4, as a team, we will all need to become comfortable with using Replit which we have never used.
As a team, the week was mostly spent on doing the design review report. We were still waiting on some hardware components to arrive so that we could finally start setting them up in public space. We think the risks that could jeopardize the success …
The week was mainly spent on writing the design review report as there was a lot to be done. I also spent time on fixing the bugs and fine tuning the YOLOv7 code so that it works to our needs. Since we haven’t been able …
This week I received the Jetson Nano 2GB and the Sainsmart IMX129 camera module that I requested from the course list. However, upon receiving the box, I realized the box did not have a power cord or SD card to boot up the OS, so I set to work looking at which ones to buy online. I settled for a 64GB SD card as the example setup videos I watched from last week used those. For the power bank, the box recommended that a 5V 3A USB-C power source be used with this board. I instead found a 5V 4A USB-C power source with an on/off switch (should be helpful in saving power while stores are closed). After verifying online that the increased current wouldn’t be a problem with the board, I added this to my shopping list.
Along with these and with the design review document coming up soon, I spent time heavily thinking of how my hardware will physically look together. Because we went with a camera module as opposed to the USB cameras from the beginning of the semester, the original orientation of the hardware has changed. The camera module must be close to the Jetson, meaning the Jetson will have to be elevated with the camera. Right now I was thinking of using reclosable fasteners to my shopping list, and perhaps have the Jetson sit on an elevated board with the camera.
Because the camera is being elevated it likely won’t be close to the power source, so I also added some adapters between USB-C and USB-A and some USB extension cables.
Finally, I wanted a compact wifi adapter for the Jetson so the physical size of the overall system wouldn’t get too large. Looking at reviews the NETGEAR AC1200 Wifi adapter was a good middle between size and transfer speed. I also picked some ethernet cables to satisfy our risk mitigation given we have problems with WIfi.
Here is a picture of the current materials list:
Additionally, since I couldn’t set up the OS like I wanted this week due to the lack of materials, I instead began downloading the softwares I will need to set up the OS, as well as start learning how to integrate the camera module. I downloaded, an SD card formatter and balenaEtcher which will be used to flash the SD card, as well as the jetson nano sd card image which holds the OS.
For interfacing the camera module, I learned that there was an open source python github called “CSI-camera” which provides code for interfacing with Jetson Nano camera modules. Moving forward I plan to use this to both format the camera module for our uses and learn how to pass the camera module into our YOLO algorithm (Some examples of facial recognition algorithms provided in the github). The video I watched where I learned this information:
Throughout this week my progress was definitely behind as I didn’t have any of the hardware I expected to have. However, I think pivoting my focus onto problems which I could solve (Getting a better idea of the physical system, preparing the shopping list for design report, beginning to learn about the camera module, etc.) I am back on schedule. The time I spent doing these tasks which I planned to do later will open back up time to set the OS like I wanted to this week.
In this next week I hope to finalize our Table of Materials and Cost with the items/softwares my other group members will need. Additionally, a lot of time will be spent on the Design Review Document, as it is due by the end of this week. If my parts were to arrive by this week I also would have liked to set up the OS this week, but it looks like they will be coming next weekend meaning I will have to complete this step after spring break.
This week, I set up the Django project and the Github repo for version control: https://github.com/santacalculus/No-Time-to-Dine-capstone/. While we had initially thought of using MongoDB Atlas to store video data, we have come up with a simpler solution on the backend side that involves storing the …
Half of the week was spent on the Design Review. There were issues related to my part specifically that were addressed in our feedbacks. First of all, there was a concern related to how we would customize the Yolov7 model to account for human objects …
The camera we decided on previously – the Sainsmart IMX219 – is not a USB camera as we had indicated previously, as has been pointed out in the feedback. However, as it is compatible with the Jetson Nano board in which our image processing algorithm will be housed, that regrettable error has been mitigated. There is also the concern that it will be difficult to keep the Margin of Error in wait time to 1-2 minutes because of a cascading effect of having about 10-30 seconds of margin of error for each individual user. So we plan on conducting an informal survey, asking individuals in La Prima if they would like to share their order and observing how long it takes for that order to be prepared, so we can use this data in our linear regression algorithm.
Our Block Diagram has also been updated as we are no longer using MongoDB Atlas and are instead using SQLite as we will be storing data on the boundary boxes and don’t need to store live video data. This change was necessary as the privacy of our users in this age of constant surveillance is important to us.
Our updated schedule is linked here
This week I researched on what is the optimal database for us to use. We considered DocumentDB by Amazon AWS as it is a popular choice as a cloud database, however, MongoDB Atlas emerged as the better choice. As discussed in our presentation last week, …