Skip to content

Latest commit

 

History

History
137 lines (93 loc) · 7.28 KB

README.md

File metadata and controls

137 lines (93 loc) · 7.28 KB
Raw

AccessiPark

Charting accessibility obstacles and accessible parking opportunities with computer vision, Google Street View, and Denver OpenData.

W210 Capstone Project - Summer 2020

by Rachael Burns, Richard Ryu, Michelle Sun, and Hong Yang

Problem Statement

Problem 1: City streets are not always accessible. According to the 2010 US Census, there are 30.6 million Americans with physical disabilities. Nearly half use an assistive aid, such as a wheelchair (3.6 million) or a cane, crutches, or walker (11.6 million). Many streets, sidewalks, and businesses in the US remain inaccessible. Parking and sidewalk accessibility fundamentally affect where and how people live their lives, yet it is challenging for anyone to determine whether a given destination is accessible. The National Council of Disability and other organizations do not have comprehensive information on the degree to which sidewalks are accessible. Where partial information relevant to accessibility exists, there are few user-friendly solutions to deliver it to those who need it most. Methods available to organizations to assess accessibility, often in-person street audits or citizen call-in reports, are inconsistent and costly.

Problem 2: Accessible parking is not easy to find. According to the Accessible Parking Coalition:

- 69% of people with disabilities have problems finding accessible parking in their communities
- 96% say parking availability is important to leading an independent life
- 70% say their decision to drive or ride is influenced by parking availability
- 62% say they would be more likely to drive or ride if parking was more available
- 52% have decided not to make a trip because of concerns about finding parking

Mission and Vision Statement

Our mission is to help people identify accessible parking opportunities and avoid accessibility obstacles.

Our vision is for people with mobility challenges to be able to lead fulfilling lives with peace of mind, armed with a plan for where to park and how to continue unobstructed to their destination.

So, what is this?

We're creating an accessible parking map that displays handicap signs and obstacles that are detected through our custom trained YOLOv5 model that's inspired by Roboflow's YOLOv5 notebook

How are we doing it?

Data

  • Used ArcGIS pro to extract all lat/long coordinates of downtown Denver
  • Leveraged Google Street API to download all images (~200k) associated with the coordinates from the extracted listed above

Downtown Denver

Custom Labelling

  • Used labelImg to draw bounding boxes for 5 classes on 2500 images

System architecture

Model Methodology

  • Inspired by Roboflow's YOLOv5 notebook, we implemented YOLOv5 on our custom labelled images

  • Applied train (70%) / test (30%) split on our dataset

  • Leveraged Google Colab and TensorFlow for both training and inference

  • mean Average Precision (mAP) was used for model evaluation

  • We won't go into much detail about YOLOv5 since this page is intended to explain how we were able to come up with AccessiPark map. For more information about YOLOv5, please visit Roboflow's YOLOv5 blog

  • For inference, we used YOLOv5 pre-trained weights to detect 5 objects on ~200,000 images of downtown Denver

Feature Engineering

The bounding boxes from the inference above are saved in a csv format. Above is an example of the csv file in pandas DF. In order to plot this on our AccessiPark map, we've listed the necessary feature engineering below:

  • Transformation

    • Split 'imgname' into 'lat', 'long', 'angle'

    • Use googlemaps reverse_geocode API to search for the corresponding zipcode for each coordinate pairs

# You will need an active googlemaps API code
# david = pd.DataFrame(columns=['new_lat', 'new_long'])

import googlemaps
gmaps = googlemaps.Client(key = 'your API key')

for i in list(range(len(david))):
    temp_result = gmaps.reverse_geocode((david.loc[i,'new_lat'], david.loc[i, 'new_long']), result_type = "postal_code")
    david.loc[i, 'zipcode'] = temp_result[0].get('address_components')[0].get('long_name')
  • Use 'xcenter' of the bounding box to determine the relative location (left, right, center) of the detected object, and then apply adjustment based on the cosine and sine transformation of the 'lat' and 'long'
def csv_par11(df, delta):
    for i in range(len(df)):
        df.loc[i, 'lat'] = float(df.loc[i, 'imgname'].split('-')[0])
        df.loc[i, 'long'] = float(df.loc[i, 'imgname'].split('-')[2]) * -1.0
        df.loc[i, 'angle'] = df.loc[i, 'imgname'].split('-')[3][:-4]

        if df.loc[i, 'xcenter'] <= 0.333333:
            df.loc[i, 'detect_loc'] = 'L'
            df.loc[i, 'new_lat'] = df.loc[i, 'lat'] + (delta * math.cos((math.pi / 180)*(float(df.loc[i, 'angle'])-90)))
            df.loc[i, 'new_long'] = df.loc[i, 'long'] + (delta * math.sin((math.pi / 180)*(float(df.loc[i, 'angle'])-90)))
        elif df.loc[i, 'xcenter'] > 0.666666:
            df.loc[i, 'detect_loc'] = 'R'
            df.loc[i, 'new_lat'] = df.loc[i, 'lat'] - (delta * math.cos((math.pi / 180)*(float(df.loc[i, 'angle']) -90)))
            df.loc[i, 'new_long'] = df.loc[i, 'long'] - (delta * math.sin((math.pi / 180)*(float(df.loc[i, 'angle']) -90)))
        else:
            df.loc[i, 'detect_loc'] = 'C'
            df.loc[i, 'new_lat'] = df.loc[i, 'lat']
            df.loc[i, 'new_long'] = df.loc[i, 'long']
    return df
  • Organize inferences into CSV file by zipcodes
def zip_df(df):
    for i in df["zipcode"].unique():
        temp_df = df[df["zipcode"] == i]
        temp_df.to_csv(i + '.csv', encoding = 'utf-8', index = False)
  • For full process from start to finish, please refer to the ETL notebook

AccessiPark Map

Come check it out at AccessiPark

Accessibility Rating

Heatmap of the downtown Denver zipcodes by number of handicap signs detected from our YOLOv5 inference