Bases: common.models.EmptyModelBase
Flickr user
if true, this user has too many bogus photos and will be ignored
flickr username
Bases: common.models.UserBase
Photograph
Default filters for views
width/height aspect ratio
name of photographer or source, if not a Flickr user
optional user description
exif data (output from jhead)
flickr photo id
flickr user that uploaded this photo
focal length in units of height (focal_pixels = height * focal_y)
Helper for svg templates
Helper for svg templates: return the font size that should be used to render an image with width 512px, when in an SVG environment that has the height scaled to 1.0 units
field of view in degrees of the longer dimension
Return a dictionary of this model containing just the fields needed for javascript rendering.
Fetch a pixel, in floating point coordinates (x and y range from 0.0 inclusive to 1.0 exclusive), at a given resolution (specified by width)
Helper for templates: return the image height when the width is 1024
Helper for templates: return the image height when the width is 512
The photograph resized to fit inside the rectangle 1024 x 2048
The photograph resized to fit inside the rectangle 200 x 400
The photograph resized to fit inside the rectangle 2048 x 4096
The photograph resized to fit inside the rectangle 300 x 600
The photograph resized to fit inside the rectangle 512 x 1024
Returns the height of image_<width>
original uploaded image (jpg format)
The photograph cropped (and resized) to fit inside the square 300 x 300
if True, this is part of the IIW (“Intrinsic Images in the Wild” paper) dataset.
if True, this is part of the IIW (“Intrinsic Images in the Wild” paper) dense dataset.
If True, this photo contains sexual content. If None, this photo has not been examined for this attribute. This field is set by admins, not workers, by visually judging the image. (this is not a limitation; we just didn’t think to make this a task until late in the project)
copyright license
the light stack that this photo is part of (most photos will not be part of one)
hash for simple duplicate detection
median intrinsic images error
If True, this photo was NOT taken with a perspective lens (e.g. fisheye). If None, this photo has not been examined for this attribute. This field is set by admins, not workers, by visually judging the image. (this is not a limitation; we just didn’t think to make this a task until late in the project)
cache of the number of intrinsic comparisons with nonzero score
cache of the number of intrinsic points (all points)
cache of the number of correct material shapes for this photo (useful optimization when sorting by this count). These values are updated by the celery task photos.tasks.update_photos_num_shapes()
cache of the number of vertices in all correct material shapes for this photo (useful optimization when sorting by this count). These values are updated by the celery task photos.tasks.update_photos_num_shapes()
Fetch the image at a given size (see the image_<width> fields)
height of image_orig
width of image_orig
True if the license exists and has publishable=True
Return a score indicating how ‘open’ the photo license is
If True, this photo is incorrectly rotated (tilt about the center). looking up or down does not count as ‘rotated’. This label is subjective; the image has to be tilted by more than 30 degrees to be labeled ‘rotated’. The label is mostly intended to capture images that are clearly 90 degrees from correct. If None, this photo has not been examined for this attribute. This field is set by admins, not workers, by visually judging the image. (this is not a limitation; we just didn’t think to make this a task until late in the project)
scene, e.g. “bathroom”, “living room”
if true, the scene category is valid; null: unknown
method used to set scene_category_correct
further from 0: more confident in assignment of scene_category_correct
Tf True, this photo does not represent what the scene really looks like from a pinhole camera. For example, it may have been visibly edited or is obviously HDR, or has low quality, high noise, excessive blur, excessive defocus, visible vignetting, long exposure effects, text overlays, timestamp overlays, black/white borders, washed out colors, sepia tone or black/white, infrared filter, very distorted tones (note that whitebalanced is a separate field), or some other effect.
If None, this photo has not been examined for this attribute. This field is set by admins, not workers, by visually judging the image. (this is not a limitation; we just didn’t think to make this a task until late in the project)
if true, this is synthetic or otherwise manually inserted for special experiments.
Sum of the length of all vanishing line segments (vanishing_lines).
This is not “vanishing lines” in the sense that it is an infinite line passing through the vanishing point. Rather, this is a list of line segments detected in the image, classified as likely passing through a certain vanishing point.
json-encoded list of groups. Each group: list of lines. Each line: x1, y1, x2, y2 normalized by width and height. The groups are stored in order of decreasing size.
Generator for all vanishing line segments as SVG path data
Return the 3D unit vector corresponding to a vanishing point (specified in normalized coordinates)
Vanishing points, json-encoded. Format: list of normalized (x, y) tuples.
Return the vanishing points as an enumerated python list
Returns vanishing points as unit vectors (converted to tuples for compatibility with templates)
Return the vanishing vectors as an enumerated python list
Return the 2D vanishing point (in normalized coordinates) corresponding to 3D vector
if true, this photo is whitebalanced; null: unknown
further from 0: more confident in assignment of whitebalanced
Bases: common.models.ResultBase
Abstract parent for photo labels
Bases: common.models.EmptyModelBase
A collection of photos that are of an identical scene and viewpoint, but with different lighting conditions
Bases: common.models.EmptyModelBase
Scene category, such as ‘kitchen’, ‘bathroom’, ‘living room’
Text description of this category
Scene category name
“Parent” category. Scene categories can be nested in a tree. Currently none are.
Bases: photos.models.PhotoLabelBase
Label indicating whether a photo’s scene category is correct.
Bases: photos.models.PhotoLabelBase
List of points in an image that should have zero chromaticity.
Median 2-norm of the a,b channel in L*a*b*, null if num_points = 0
Returns the points with the x coordinate scaled to have the correct aspect ratio
number of points
photo for this whitebalance label
Point format: x1,y1,x2,y2 as a fraction of width,height