Mapillary is doing a great job, I hope be contributing soon to this project which I onyl recently came across.
One of my observations is that Mapillary doesn’t always detect sub-signs. E.g.: A parking sign is detected but it doesn’t detect the two-wheeler motor cycle sign right under the P sign.
So it effectively only detects only 1 of the 2 street signs.
The result is the conclusion we can draw from mapillary is different from the reality:
Only ‘P’ detected: parking
but ‘P’ with subsign two-wheeler motor cycle : parking for motor-cycles only.
There are numerous such examples, like, 1. no-entry (well detected) vs no-entry other than bicycles (bicycles not detected in sub-sign)… 2. Speed limits that apply only to a particular types of vehicles 3. Speed limits that apply to all vehicles taking the exit… etc.
So my question is:
- Why are sub -signs more often not detected than detected ?
- The un-detected signs (sub-signs that caught my interest) are a result of the sign not being included in the official image recognition traning?
Thanks for your clarification