top of page

Extracting Insights from Manhattan's Street View Images - Computer Vision Project Help


Extracting insights from street view images can provide valuable information about a city's urban environment. In this project, we are tasked with generating a map of Manhattan, New York City, based on provided street view images. By applying computer vision techniques, we aim to extract meaningful data from these images and visualize it on a map. This map will showcase various attributes such as the distribution of pedestrians, categories and numbers of vehicles, sky view area, and greenery area. Additionally, we have the opportunity to incorporate other datasets to enhance the value of the map, such as estimating green space exposure by adding population data or creating a novel urban design/management index by incorporating points of interest (POI) or other relevant information. The possibilities are extensive, and we encourage thinking outside the box to push the boundaries of what can be achieved. By showcasing novelty, quality, and value in our map, we can earn a high score for this project.

Task Description

The task involves working with street view images of Manhattan, NYC, and their corresponding geolocations. Using computer vision techniques, we will extract interesting information from these images, which could include the distribution of pedestrians, types of vehicles, sky view area, greenery area, and more. Based on the extracted attributes, we will generate a map that visualizes the information effectively. To add value to our map, we are encouraged to incorporate other datasets. For example, we can estimate green space exposure by integrating population data or create a novel urban design/management index by including points of interest (POI) or other relevant factors. The goal is to push our creativity and explore the full potential of the provided street view images. Extensive research articles can serve as inspiration and guidance throughout the project.

Data Description

Dataset comprising 29,540 street view images of Manhattan sourced from the Google Maps Platform. The images were sampled every 30 meters from the New York road map, covering a significant portion of Manhattan. The geolocations of the images are embedded within the image file names, following the format "latitude_longitude.jpg" (e.g., 40.6885395196_-74.02375583308243.jpg). The dataset can be accessed from the following link: [Insert link to dataset].

Computer Vision Techniques

To extract meaningful information from the street view images, we are required to employ at least one computer vision technique. Several options are available, including object detection, semantic segmentation, image recognition, instance segmentation, and other interesting approaches. We are encouraged to explore codes and methods beyond what was covered in class, allowing us to showcase our creativity and expertise in computer vision.


For reference and to facilitate code replication, an Explanation.ipynb file has been provided. This file contains examples and explanations of techniques used to generate the map. We are free to leverage the code and adapt it to our specific needs.

Output The final deliverables for this project consist of:

a. A PDF file comprising:

  1. One A4 page map that presents the visualized information extracted from the street view images.

  2. Additional pages explaining the process behind generating the map, including details such as the target objects, computer vision techniques employed, and the steps taken to visualize and enhance the map's aesthetics.

  3. A discussion of the findings discovered through the map, providing insights and analysis based on the extracted information.

b. A zip file containing:

  • The codes used for the project, extracted from Google Colab or any other coding platform.

  • Any extra data that was used, properly cleaned and formatted.

  • The source(s) of the additional data should be clearly specified.


  1. To enhance the map's visual appeal, it is advisable to incorporate fishnets, grids, hexagons, or other techniques. Interpolation, extrapolation, or inverse distance weighting (IDW) can also be utilized, as long as they are reasonable and contribute to the map's quality.

  2. Preprocessing techniques may be necessary to improve the panoramic street view images, as they might be distorted compared to standard images. It is also encouraged to leverage state-of-the-art computer vision techniques available on platforms like GitHub to achieve the best possible results for the map.

If you require a solution for this project or need assistance in enhancing your street view image analysis, our team at CodersArts is well-equipped to support you. With our expertise in computer vision and data analysis, we can help you extract valuable insights and generate visually appealing maps. Please feel free to contact us via email or through our website. Let us help you unlock the full potential of your street view image data and create impactful solutions for your urban planning or management needs.

bottom of page