FINAL PROJECT
Final project topics: You must identify a topic that interests you and agreed by the instructor. See the list of possible topics for inspirations. For topics, the only requirement is that the topic must be civil engineering, environmental engineering, or engineering mechanics related. You are welcome to explore applications of machine learning in the wider realm of these fields (e.g., structural, geotechnical, environmental, traffic, urban). The evaluation of the final project will not be affected by the topic selected.
List of possible topics (note that for all these topics, I can give you some general ideas of what specific tasks are out there and can be solved by machine learning techniques, but you are still responsible to find the data sourcefor specific tasks on your own):
• Construction worker safety
• Structural health monitoring
• High performance building monitoring and control
• Building cluster energy consumption and demand response
• Construction robotics (e.g., navigation, object detection, assembly, human-robot collaboration)
Final project team formation: you are expected to form. teams of two for the final project. Single person team is also allowed.
Final project consultation: you are welcome to book a 30-min time slot with me (feel free to email me) and discuss any project ideas.
Final project check-in: To facilitate the completion of the final project, we will have a final project check-in towards the beginning of the second half of the semester. For the check-in, you are expected to submit a maximum 2-page report outlining: (1) the main objective of the final project (2 points), and (2) existing/possible data sources (note that if there is no existing data source, a mechanism for collecting data needed for the project is required) (3 points).
The final deliverable of the final project will be: (1) a report of the findings of the final project, and (2) a 10-minute presentation of the final result/product of the final project. You are also encouraged, but not required, to publish your work on open online repositories (e.g., GitHub).
Final Report Guidelines:
The final report should be maximum 10 pages in length, single space, 12 pt. Times New Roman, default margin. Include the following sections:
Introduction (1-2 pages, 2 points): First introduce the significance of the problem. For example, if you are predicting energy consumption of buildings, you can mention facts such as buildings account for 45% of the energy consumed in the U.S. every year. Next, formulate the problem. How do you formulate your problem into a machine learning problem? What is the input, output? Where did you get the data?
Background (1-2 pages, 2 points): what is the state-of-the-art model for the problem you are trying to solve? For example, if you are solving a hardhat detection problem, what is the best performing model for the problem hardhat detection? Why did you choose the model you are using? Cite the papers/sources you are referencing and give a succinct summary of their performance metrics and results.
Methodology (2-3 pages, 8 points): lay out the model you used, the hyperparameter of the model, and the overall structure of model (e.g., if you are using CNN, what is the CNN structure, how many parameters, what are the layers, etc.). A figure depicting the overall work flow of the methodology is helpful. What did you do in each of the steps for this project (e.g., data collection, data cleaning, model architecture used, validation methods used, etc.)
Metrics (1 page, 2 points): give definitions/equations of the metric you used.
Results (3-4 pages, 4 points): show the results for your models, using the metrics you defined in the last section.
Limitations (1 page, 2 points): what are the limitations for this project? How would you improve the project if you have (e.g., high performance computers, larger datasets, better ML models, more time, etc.)?
Reference (does not count towards the total number of pages): give a list of references.
10-Minute Final Presentation Guidelines:
This should be a presentation that captures the main idea and results of the project. It should show the most important and coolest parts of the project, with narration to explain how it works. Think of it as a 10-minute thesis presentation. The following criteria will be used for grading:
• Did every member on the team present (0.5 point)?
• Did the presenter properly motivate the audience regarding the problem being solved (1 point)?
• How is this problem currently being addressed in the industry, in academia (1.5 points)?
• Was the problem formulation clearly stated (1 point)?
• Was the methodology clearly explained (2.5 points)? (What model, why this model? What architecture, why this architecture? What other models did you use as comparisons?)
• Were the results clearly explained with proper figures, tables (2 points)?
• Did the presenters provide a concise take home message for this project (0.5 point)?
• Did the presenter properly acknowledge the limitations of this project and identified future directions (1 point)?