Author: shayang

Progress Made

My main focus this week was with prepping and structuring the CV algorithm. I’ve been coding the training algorithm using the TensorFlow library and mainly debugging on smaller datasets (so I can run locally). 

I also helped the team finalize our parts order, which was sent out!

Progress

On Schedule! 🙂

Deliverables for Next Week

• Procure AWS credits and find out how to train model using AWS
• Train CV model on COCO dataset using TensorFlow
• Hopefully be able to do some validation with the photos I took too

Progress Made

My main focus this week was with data collection. I continued to amass photo datasets from online as well as take photos of my own. I made sure to visit a variety of intersections with just traffic lights (including Ellsworth/Morewood, Ellsworth/Amberson, Ellsworth/Aiken, Fifth/Amberson, Fifth/Wilkins). We had some rare sunny weather, so I seized the opportunity to snap some pics then. I also got pics during the normal Pittsburgh overcast, in some light rain, and at night. I took photos at various angles and approaches, as well as with different light stages.
Link to photos so far

Progress

On Schedule! 🙂

Deliverables for Next Week

• Can always get more photos with more varied weather. (Heavier rain in the forecast!)
• Set up CV architecture and reacquaint myself with OpenCV

This week, I did the following:

• Project Proposal refinement and Scoping
  • Worked with the Team as well as Dr. Sullivan and Rashmi to continue to narrow the scope of our MVP, starting from a short list of possible features to deciding on telling the user when to cross at a specific “ElMo” type intersection
• Literature Review / Data Collection
  • Helped identify traffic datasets for CV
    • API with Google Street View for images of specific intersection
    • Microsoft et. al. COCO dataset with tagged photos of traffic lights
• Prep work for circuitry assembly
  • Selection of speaker for user interface

Perceived Team Progress:

On Track 🙂

Deliverables next week:

• Data Collection
  • Collect pictures of “ElMo” intersections (incl. Ellsworth/Morewood, Ellsworth/Amberson, Fifth/Amberson) at various angles/vantage points, lighting conditions, weather conditions
  • Tagging the traffic lights in said photos
• Project Requirements
  • Finalize, through literature review and similar past projects, the quantitative benchmarks that are required by the user (i.e. data rate, power consumption, accuracy)
• Device Components
  • Finalize parts list with team and order