Blog Posts

Emotional Intelligence Increased!

This week, the Fantastic 5 had a presentation about emotional intelligence by Ian Baldwin, Director of Services from Infotech Inc. (Pictured below!) We learned a lot of valuable information about working in teams and using emotions to guide cognitive thinking.

In regards to our AI Image Generator project, we are currently working on our prototype to get ready for Prototype Inspection Day! We also found out that most machine learning frameworks accept image-text pairs in the form of a CSV file, where each row contains the path to an image and its corresponding text description. It is also possible to have multiple images associated with a single text description, which will be helpful when training with multiple jersey product views.

PDR Presented!

This week the Fantastic 5 presented our preliminary design report to our sponsor Fanatics on October 25. The presentation went well! The team at Fanatics seemed excited about our proposal and were giving us ideas on how to expand the functionality of the AI. One feedback we received was to show an example of how the AI would work using a random AI model from the internet.

Now that we presented our PDR, the PDR will be sent out to our sponsor to be signed off!

One of our team members, Sara, learned about federated learning and she is hoping to apply it to our software. Federated learning is employed to ensure no data is leaked during the training process of the ML models.

PDR Presentation Practice

Peer Review Presentation

This week, our team got to work on our presentation for the preliminary design report. Our slides came out fantastic and we had the opportunity to practice presenting in front of our peers and receive feedback. The team did great during the presentation, but some of the feedback we received was mostly to practice more to memorize the information we need to present. For the upcoming week, we are excited to present our PDR to Fanatics on Wednesday and show off all our hard work and research!

Stable Diffusion XL

Upon further research, our team discovered this version of stable diffusion that has more features. It improves on the original in three different ways:

  • The UNet is 3x larger and SDXL combines a second text encoder (OpenCLIP ViT-bigG/14) with the original text encoder to significantly increase the number of parameters
  • Introduces size and crop-conditioning to preserve training data from being discarded and gain more control over how a generated image should be cropped
  • Introduces a two-stage model process; the base model (can also be run as a standalone model) generates an image as an input to the refiner model which adds additional high-quality detail

Plans For Next Week

  • Update our software architecture to embed into Fanatics architecture
  • Improve on our presentation based on feedback we received and prepare
  • Continue learning about Stable Diffusion XL

Going Deeper into AI

Hello everyone, this week our team has been hard at work learning more about AI works. We learned the difference between unconditional vs conditional generative AI, generative vs discriminative techniques, and how to build a rough ML model script using Python and libraries found on Hugging Face. It is exciting times ahead for us! We also received some test data from Fanatics that we can use to train an AI model, so in the coming weeks we are going to be trying that.

Hugging Face, website where a lot of trained data models are.

Deliverables

The team also worked on a preliminary product architecture. We are waiting to hear back about how Fanatic’s backend structure works to improve on it. We also have been working on our product design report and finishing up the draft. We are presenting next Tuesday to the class and getting valuable feedback on how to improve our presentation for the real presentation to our sponsor Fanatics!

White Belt Certified!

The Past Weekend

Last Saturday, we went to a white belt workshop hosted by FPL where we learned how to create diagrams to improve on a process and tackle problems. Above is the picture of our team with the diagram we created to improve car sales!

Work This Week

Preliminary Design Specifications

We updated our Technical Performance Measures to incorporate feedback we received to reflect the business impact of our solution. So now we are measuring the cost to generate images compared to the cost of photographing images. We also added the measure of how may product display pages are able to be published.

Project Roadmap

We completed our project roadmap for the rest of fall. We will complete spring by the end of the semester. Major dates approaching are October 23-27, when we have to present our Preliminary Design Report.

Next Week

We will be working on our architecture diagram which is part of our PDR presentation. We will also be working on the draft for our PDR which has to be approved by Fanatics. We also will be getting the data necessary to train our AI model.

Planning for the Future

Image by https://www.freepik.com/free-vector/hand-drawn-business-planning-with-calendar_19948900.htm#query=planning%20graphic&position=48&from_view=keyword&track=ais

Research

Another week has flown by and the Fantastic Five have been hard at work researching viable methods to generate AI images. We found two potential ways we could accomplish the goal: using either OpenAI’s API or creating our own model from scratch and using stable diffusion. We played around with stable diffusion and created some example images and found it to be straightforward when using an already existing model.

Example of jersey we will have to generate

After talking to our liaison, it looks like using stable diffusion would be the preferable method, but that also means training our own model using data that we preprocess and prepare. We have our work cut out for us, but we are excited to learn more about training our own AI model!

Preparation

This week, we also finished writing our design specifications and came up with our customer needs. With that finished, we are planning on polishing it to present it to our sponsor Fanatics. We are also preparing a road map for the rest of the semester. This will allow us to stay organized and on track with our project. Eventually, we hope to also expand the roadmap to encompass the spring semester as well. Next week, we will start working on the preliminary design ideas.

The Fantastic Five!

Our Logo

Hello again everyone, this is the end of the second week working on our project! This week we finalized our team name, the Fantastic Five, as well as our team logo!

We are called the Fantastic 5 because fantastic is a play on Fanatic and there are five on us on the team and we also happen to be team 5.

For our logo, we incorporated a flag because the Fanatics logo features a flag. We also met with the Fanatics liaison and got clarity for our scope of work and did more research on AI and what API’s we can use or models we can train.

As a team, we started working on lists of project requirements, customer needs, and specifications of the product design.

For the upcoming week, our plan is to keep researching with the clarifications we got and present some of our ideas to our liaison. We are starting our research by looking into an AI API that can accept images and is able to generate images.

Tagged as: , ,

Week 1

Hi everyone, welcome to our blog! We are a team working for Fanatics during the academic year in the IPPD program. We will document our experience working on our project and our progress throughout the year.

This week, we found out about our project and got to meet each other for the first time. We also started doing research on our project and coming up with ideas for our team’s name and logo. We shared our availabilities with each other and planned out how each week will look in terms of meeting up and working together. We met up with our coach and introduced ourselves and sent an email to our liaison to check his availability.

During the next week, we will be meeting with our liaison and finding out more details about our project, such as figuring out project scope and deadlines.