SYSTEM AND METHOD FOR HOME FAÇADE FEATURES DETECTION IN A GEOGRAPHIC REGION
A method including receiving a first set of one or more street level images of houses into a Machine Learned Model trained with a second set of street level images with one or more exterior features of the houses labeled, identifying the one or more exterior features in the first set of street level images by way of the Machine Learned Model, and quantifying and outputting the counts and/or two-dimensional areas for each of the identified exterior features in the first set of one or more street level images is described. Non-transitory, computer-readable storage media having instructions for executing the method steps by one or more processors as well as computer or computer systems capable of performing the method steps are also described.
1 . A method, comprising:
receiving one or more raw images of houses, each image taken from a viewpoint from a street adjacent to each house in the raw images, thereby resulting in raw street level images of houses;
cropping each raw street level image of each house based on an overhead view or representation of the house to remove portions of adjacent houses in the raw street level images thereby producing cropped street level images of each house;
inputting the cropped street level images into a Machine Learned Model comprising a Convolutional Neural Network trained with a training set of street level images with one or more exterior features of the houses labeled;
identifying the one or more exterior features in the cropped street level images by way of the Machine Learned Model;
quantifying and outputting, in integers such as 0, 1, 2, or 3, counts of each of the identified exterior features in the cropped street level images;
converting pixels of the identified exterior features in the cropped street level images into area dimensions; and
quantifying and outputting, in area units such as square feet or square meters, a two-dimensional area for each of the identified exterior features in the cropped images.
2 - 4 . (canceled)
5 . The method of claim 1 , wherein cropping the raw street level images comprises:
identifying corners of each house in the raw street level image by:
determining an angular extent of the house from a nearest point on the street from the overhead view or representation of the house; and
translating distance and angles from the determining step to pixel values in the raw street level image; and
calculating portions of the raw street level image in excess of pixel values representing the house; and
removing the portions of the raw street level in excess of pixel values representing the house.
6 . The method of claim 1 , further comprising:
receiving a geographic area of interest;
storing the geographic area of interest;
retrieving addresses of houses within the geographic area of interest;
retrieving the one or more raw street level images of individual houses based on the retrieved addresses.
7 . The method of claim 6 , wherein the geographic area of interest is received as input on a graphical user interface, the input comprising a boundary on a map.
8 . The method of claim 6 , wherein the geographical area of interest is received as input on a graphical user interface, the input comprising one or more of street, block, neighborhood, town, city, county, state, metropolitan statistical area (MSA), core-based statistical area (CBSA), five-digit ZIP Code, nine-digit ZIP Code, geographic coordinates (latitude and longitude), and tax parcel number.
9 . The method of claim 1 , wherein the outputting comprises text representing characteristics of each cropped street level image, the text having information comprising:
a unique alphanumeric identifier;
street address;
geographic location;
date of image acquisition; and
image URL.
10 . The method of claim 1 , wherein the outputting comprises information representing characteristics of each cropped street level image, wherein the information further comprises counts of individual home exterior features identified in the image.
11 . The method of claim 1 , wherein the outputting comprises information representing characteristics of each cropped street level image, wherein the information further comprises calculated areas of individual home exterior features of the image.
12 . The method of claim 9 , wherein the text is formatted and standardized for input into a geographic information system application.
13 . The method of claim 1 , wherein the counts and/or two-dimensional areas for each of the identified exterior features for the cropped street level images are outputted as a bar chart.
14 . The method of claim 1 , wherein the counts and/or two-dimensional areas for each of the identified exterior features for the cropped street level images are outputted as a heat map.
15 . The method of claim 1 , wherein the counts and/or two-dimensional areas for each of the identified exterior features for the cropped street level images are converted to a percentage of total counts or total area for all of the identified exterior features and outputted as a pie chart.
16 . The method of claim 1 , wherein the identified exterior features are chosen from windows, doors, garage doors, roofs, horizontal façades, vertical façades, stucco façades, brick façades, shakes façades, and stone façades.
17 . The method of claim 16 , wherein the identified exterior features are identified as requiring repair or replacement.
18 . The method of claim 1 , wherein the identified exterior features are chosen from windows, doors, garage doors, roofs, horizontal façades, vertical façades, stucco façades, brick façades, shakes façades, and stone façades.
19 . A method of home exterior feature detection and analysis, wherein a geographic area of interest and home exterior features of interest are received on a graphical user interface as input, the method comprising:
storing the geographic area of interest;
retrieving addresses of all houses within the geographic area of interest;
retrieving and storing raw images of houses taken from a viewpoint from a street adjacent to each house for each of the retrieved addresses, thereby resulting in raw street level images of houses;
cropping the raw street level images based on an overhead view or representation of each house to remove portions of adjacent houses in the images thereby producing cropped images of each house;
storing the cropped images;
providing the stored cropped images to a Machine Learned Model trained with a training set of street level images of with one or more exterior features labeled;
identifying the home exterior features of interest in the cropped images by way of the Machine Learned Model;
quantifying and outputting, in integers such as 0, 1, 2, or 3, counts of each of the identified exterior features in the cropped street level images;
converting pixels of the identified exterior features in the cropped street level images into area dimensions; and
quantifying and outputting, in area units such as square feet or square meters, a two-dimensional area for each of the identified exterior features in the cropped street level images.
20 . The method of claim 19 , wherein cropping the raw street level images comprises:
identifying corners of each house in the raw street level image by:
determining an angular extent of the house from a nearest point on the street from the overhead view or representation of the house; and
translating distance and angles from the determining step to pixel values in the raw street level image; and
calculating portions of the raw street level image in excess of pixel values representing the house; and
removing the portions of the raw street level image in excess of pixel values representing the house.