7.13.0

[misaugstad]

• Labels no longer move immediately after you place them Labels move after placement #1764 (PR Fixes bug where labels moved after placing them #3194)
• Fixes data in APIs used for CV (important details below) image_width and image_height incorrect in database #2485 (PRs Fixes bug where labels moved after placing them #3194 Fixes old sv_image_x/y data, renames some db columns #3199)
• Some data in APIs have been renamed for clarity (details below) Rename some db columns / API props to clarify when we refer to panos #3165 (PR Fixes old sv_image_x/y data, renames some db columns #3199)
• Our Mapbox API key is now hidden Hide Mapbox API key in an env variable #3195 (PR Adds Mapbox API key as an environment variable #3197)
• Columns in our database have been moved around for clarity Record some pano metadata in a new table #3171 (PR Fixes old sv_image_x/y data, renames some db columns #3199)

Below are the specific changes to know for anyone who uses our /adminapi/panos, /adminapi/labels/panoid, /adminapi/labels/cvMetadata, /v2/access/score/streets, /v2/access/score/neighborhoods, /v2/access/attributes, or /v2/access/attributesWithLabels APIs.

Note: only GeoJSON and CSV APIs are affected, nothing has changed for Shapefiles due to a 10 character max for attribute names.

• /adminapi/panos and /adminapi/labels/panoid
  • image_width renamed to width
  • image_height renamed to height
  • photographer_heading renamed to camera_heading
  • photographer_pitch renamed to camera_pitch
• /adminapi/labels/cvMetadata
  • image_width renamed to pano_width
  • image_height renamed to pano_height
  • photographer_heading renamed to camera_heading
  • photographer_pitch renamed to camera_pitch
  • sv_image_x renamed to pano_x
  • sv_image_y renamed to pano_y
• /v2/access/score/streets (GeoJSON & CSV only)
  • avg_image_date renamed to avg_image_capture_date
• /v2/access/score/neighborhoods (GeoJSON & CSV only)
  • avg_image_date renamed to avg_image_capture_date
• /v2/access/attributes (GeoJSON & CSV only)
  • avg_image_date renamed to avg_image_capture_date
• /v2/access/attributesWithLabels (GeoJSON & CSV only)
  • image_date renamed to image_capture_date

Furthermore, if you are using our APIs for computer vision work (likely with some of the /adminapi/panos, /adminapi/labels/panoid, or /adminapi/labels/cvMetadata APIs), then you will also need to know about the following updates (roughly in order of importance):

1. We have updated sv_image_y (now called pano_y) to range from 0 to pano_height, where 0 is at the top of the image. In the past, we were outputting the y-pixel offset from the center of the image. Instead of having to compute pano_y = pano_height / 2 - sv_image_y, you should now just use pano_y.
2. Most of the errors in sv_image_x (now called pano_x) data have been fixed, so you should remove any post-hoc fixes in your code for them. The following have been fixed (and have been fixed retroactively for all data):
   1. We now take into account the width of the GSV image, so you no longer need to scale based on image size yourselves. This also fixes some pano_x values that were wildly incorrect in edge cases.
   2. We now take into account the photographer_heading (now called camera_heading) of the GSV image. You should remove corrections based on photographer_heading that you are doing yourself.
   3. We fixed a small source of error (max 45 pixels) that came from truncating the user's heading to an integer when calculating pano_x.
3. Similarly, most of the errors in sv_image_y (now called pano_y) data have been fixed, so you should remove any post-hoc fixes in your code for them. Unlike with pano_x, we still have a source of error in pano_y that we haven't been able to resolve yet -- there seems to be some level of error in the pitch that is different at different headings, which is exacerbated by hilly streets. But the following has been fixed (and has been fixed retroactively for all data):
   1. We now take into account the hieght of the GSV image, so you no longer need to scale based on image size yourselves.
   2. We fixed a small source of error (max 45 pixels) that came from truncating the user's pitch to an integer when calculating pano_y.