This week the team gave the PDR presentation on site at CAE USA. We had a great time meeting the liaisons we have been working with in person. We also got some good feedback and comments from other personnel associated with CAE as well. The overall sentiment seemed to be that we are making good progress and the team hopes to keep up with the same pace in the future!
With the PDR presentation being complete, the team now awaits for the official sign off on the proposed preliminary design from CAE. Besides the PDR the team continues to work on data collection of the plugs. We are just a little under halfway complete with the initial data set. While the data collection continues, the next milestone the team is looking forward to is prototype inspection day.
This week the team presented our draft PDR presentation to two other teams. We received great feedback that we will incorporate into our presentation before presenting to CAE next Friday. After our initial presentation, we showed our liaisons the current presentation and received additional feedback that we will take into consideration when making changes. We also finished the draft PDR report and had our coach look over it to give critiques. We plan to update the report based on the feedback so CAE can print copies of the report for the presentation if they desire.
During this upcoming week, we plan to make all the necessary changes to our PDR report and presentation and prepare for the in-person presentation. We are focusing much of our time on proper data collection so we can start efficiently training our CNN. The goal is to work on CNN architecture and techniques in parallel with data collection but we cannot start properly training without a complete dataset.
This week the team focused on preparing for the PDR Report draft presentation next Tuesday. This includes working on the Software Architecture design and outlining critical sub functions that will be included in our PDR Report.
We also were able to finalize a camera mount and get it on the probes to start mass data collection. This is a big part of our project that will take multiple days to complete. Going forward we will be able to start optimizing our current CNN architecture on relevant data to hopefully get a live demo working for before the end of this semester.
Next week we will have discussions with our liaisons on how to deal with different variables that we have encountered in our data collection. We hope these discussions will help speed up the process and completion of data collection so we can solely focus on optimizing the preprocessing and CNN.
This week the team finalized the Project Roadmap and Project First Month Report and submitted them. We have then moved on to designing the system architecture for the Preliminary System/Product Architecture assignment.
Outside of deliverables we have made good progress on the data preprocessing and gave a short demo to our liaisons. They gave great feedback on changes that could be made and emphasized that they believe what we have already is going in the right direction. The team also gave updates on the CNN and how we were successful in training a model on the preliminary data that has been gathered. There is still lots of work to do in regards to the CNN and that is what we are focusing our efforts on improving in the coming weeks.
Lastly, the team has coordinated a date to present the PDR onsite at CAE! Friday, October 22nd we will be traveling to Tampa to give our PDR presentation in person to our liaison’s and other CAE staff interested in the project. We are excited to meet everyone in person and are looking forward to presenting.
This week the team started working on the convolutional neural network (CNN) that will be trained to identify LRU’s and the pins that are being contacted. We have started off by just training the CNN on the mnist dataset (that is provided with tensorflow) to better understand how a CNN works. Moving forward we will be using the computer vision system to capture the data that will then be preprocessed and passed into the CNN. The goal is to develop the preprocessing system and the CNN in parallel to maximize productivity and progress.
Apart from development, the team has been working on the Project Roadmap and the Project First Month Report. We also have started our risk assessment by identifying potential points of failure in the development/structure of our project that may require changes to the Project Roadmap.
This week our team standardized the development environment we will be working in. This included properly setting up WSL2, Visual Studio, and all other dependencies we will be using at the start of this project. From there we were able to create a basic OpenCV program that used edge detection algorithms to identify pins from pictures of canon plugs that were found online. We received great feedback from our liaisons when demoing this program and are excited to keep developing.
On that note, we received all the equipment needed to continue development. This includes canon plugs from CAE, the endoscopic camera, and multimeter probes. The next step is to continue developing our computer vision (with OpenCV) system for data collection and to start implementing the convolutional neural network. We are brainstorming ideas of how to implement this system with our continuing work on the Concept Search and Systematic Exploration document.
Our team started off the week by finalizing the endoscope camera that will be used during the development of our computer vision system. This decision is vital to the progress of our project because no data collection can happen until we have the camera.
Also, our team was successful in finalizing the team logo and received approval from our liaisons during the weekly Wednesday meeting we have with them. During that same meeting, the team presented our first draft of the Product Design Specifications and received helpful feedback and tips on how to improve the PDS.
Lastly, the team has started to get familiar with OpenCV, the open-source framework that will be used to develop the computer vision system. Our goal is to create a basic data collection system by next week so we can start collecting a data set as early as possible.
Hello and welcome! This is the first week for team MultiVision. During this week, the team met with their coach and liaison to understand what they are tasked with. For the next two semesters, team MultiVision will be working with CAE USA Inc. in order to create a smart training multimeter. The team will mainly be creating a computer vision system to identify different cannon plugs for the smart multimeter to utilize.
For the up coming week, the team is excitedly researching Keras and Tensorflow to ready ourselves for the deep learning network we’ll be creating. At the same time, we are also looking into purchasing the camera and multimeter needed for the project.