1 of 32

Computer Vision in Post Occupancy Evaluation

Kaicheng Zhuang | SMArchS Urbanism ‘25

Kareem El-Sisi| MCP + CS SM ‘25

Zhi Ray Wang | SMArchS Urbanism ‘25

Digital City Design Workshop | course code 11.320

Urban Visual Intelligence

MIT Department of Urban Studies and Planning

Spring 2024 | February 9 - May 12 | Fridays 9am-12pm EST

Instructors:

Carlo Ratti, Professor of the Practice ratti@mit.edu

Claire Gorman clairego@mit.edu

2 of 32

Post Occupancy Evaluation

Post-occupancy evaluation (POE) refers to the process of evaluating buildings or public spaces after they have been occupied, focusing on the performance and effectiveness of the space from the perspectives of occupants. It examines various factors such as environmental conditions, space usage, user satisfaction, and building operations to identify successes and areas for improvement.

POE is important because it provides feedback on whether a building meets its intended purposes and the needs of its users, informs future building projects by identifying design strengths and weaknesses, and can lead to adjustments that improve building performance, occupant comfort, and overall satisfaction. This process is essential for creating more functional, efficient, and user-centered spaces.

3 of 32

Intended function vs. Actual usage of space

4 of 32

POE in Urban Design

In urban design, POE assesses how well public spaces, such as parks, squares, streets, and urban precincts, meet the needs and expectations of their users. It evaluates aspects like usability, accessibility, safety, comfort, social interaction, and environmental quality.

By gathering insights on how public spaces are used and experienced by the community, urban designers and planners can make informed decisions to improve existing spaces and guide the development of future projects. POE in urban and public space design helps create more vibrant, inclusive, and functional environments that better serve their communities.

5 of 32

Computer Vision in Post Occupancy Evaluation

[Biases in]

6 of 32

Biases in POE Surveys

1. Selection Bias: Findings might favor certain groups’ needs and experiences over others.

2. Response Bias: Participants might provide socially desirable responses or feedback influenced by their current mood or specific events, rather than their overall, long-term experience with the space.

3. Observer Bias: The evaluators' perspectives, expectations, and knowledge can influence how data are collected, interpreted, and reported, possibly skewing results towards preconceived notions.

4. Temporal Bias: The timing of the evaluation can affect results. For instance, a newly opened space might receive overly positive reviews due to its novelty, or specific events at the time of evaluation could temporarily alter the use and perception of the space.

5. Methodological Bias: The tools and methods used for POE (surveys, interviews, observations, etc.) might be more suitable for capturing certain types of data or feedback, potentially overlooking aspects of space use or occupant experience that are less tangible or harder to quantify.

6. Expectation Bias: Pre-existing expectations about the building or space's performance can influence both how participants perceive it and how evaluators interpret findings.

7 of 32

William H. Whyte

8 of 32

Vs.

Subjective Response

Subjective Observation

9 of 32

Vs.

Objective Observation

Long Term Study, Larger Scale

Subjective Observation

Short Term Footage, Specific Location

10 of 32

CV in Addressing the POE Bias

1. Selection Bias: Findings might favor certain groups’ needs and experiences over others.

2. Response Bias: Participants might provide socially desirable responses or feedback influenced by their current mood or specific events, rather than their overall, long-term experience with the space.

3. Observer Bias: The evaluators' perspectives, expectations, and knowledge can influence how data are collected, interpreted, and reported, possibly skewing results towards preconceived notions.

4. Temporal Bias: The timing of the evaluation can affect results. For instance, a newly opened space might receive overly positive reviews due to its novelty, or specific events at the time of evaluation could temporarily alter the use and perception of the space.

5. Methodological Bias: The tools and methods used for POE (surveys, interviews, observations, etc.) might be more suitable for capturing certain types of data or feedback, potentially overlooking aspects of space use or occupant experience that are less tangible or harder to quantify.

6. Expectation Bias: Pre-existing expectations about the building or space's performance can influence both how participants perceive it and how evaluators interpret findings.

11 of 32

Criterias

Whether people interact with the design or not?

Avg. number of people staying on site.

People’s expression when interacting with the design.

Calculate the ratio of People staying on site in groups, and people staying on site alone. Comparing with surroundings.

People’s walk speed on site, vs. avg. walking speed in the surroundings.

Percentage of people gather on site (Merge together from different direction)

Race, Gender, Identity

1

2

3

4

5

6

7

12 of 32

Criterias

Criterias proving people like the design or not?

What does the design achieve, and is that what the designer intentionally want?

Whether people interact with the design or not?

Avg. number of people staying on site.

People’s expression when interacting with the design.

Calculate the ratio of People staying on site in groups, and people staying on site alone. Comparing with surroundings.

People’s walk speed on site, vs. avg. walking speed in the surroundings.

Percentage of people gather on site (Merge together from different direction)

Race, Gender, Identity

×

1

2

3

4

5

6

7

13 of 32

Vs.

Whether people interact with the design or not?

Avg. number of people staying on site.

People’s expression when interacting with the design.

Calculate the ratio of People staying on site in groups, and people staying on site alone. Comparing with surroundings

People’s walk speed on site, vs. avg. walking speed in the surroundings.

Whether people interact with the design or not?

Avg. number of people staying on site.

People’s expression when interacting with the design.

Calculate the ratio of People staying on site in groups, and people staying on site alone. Comparing with surroundings

People’s walk speed on site, vs. avg. walking speed in the surroundings.

1

2

3

4

5

BEFORE

1

2

3

4

5

AFTER

14 of 32

Analyzing Usage Through Heat map

Heatmaps generated with Ultralytics YOLOv8 transforms tracking data into a color-coded matrix. Warmer hues indicate higher intensities and cooler tones signify lower values.

15 of 32

Methodology

  1. Take Videos at different public spaces at MIT campus, Subsampling Frames

  • Model Setup: The YOLO model detects pedestrians in each frame, while Deep-SORT maintains the continuity of pedestrians across frames.

  • Model Training + Fine-Tune: Trained the YOLO model using the COCO dataset. To improve the model’s accuracy, we fine-tuned it with additional labeled images.

  • Base on tracking data, we produce heatmaps

  • Normalize and colorize the cumulative heatmap that was gathered from each frame

16 of 32

Analyzing Usage Through Heat map

17 of 32

Analyzing Usage Through Heat map

18 of 32

19 of 32

20 of 32

21 of 32

22 of 32

23 of 32

24 of 32

25 of 32

26 of 32

27 of 32

28 of 32

29 of 32

30 of 32

Privacy Protection

31 of 32

Application

  • Understanding the limitation of existing response based Post Occupancy Evaluation
  • Introducing an objectively observing based POE model for future urban and architectural project.

  • Projecting the outcome of this research to larger scale:

Seeking possibilities of integrating computer vision related examine process into plannings and development, such as TOD. Comparing large scale surveillance footage (before and after), to determine whether a project is successful or not.

32 of 32