New Vehicle Recognition Feature Needs Help

[KachinaShadow]

I activated Vehicle Recognition a couple of days ago and I'm not impressed... 
Out of seven frames of a silver car, 4 were the same identifiable car, 2 were another car and 1 was yet another...which might be OK if you could deselect the three that shouldn't be included with the 4 frames that I could identify.

This has happened on numerous vehicles including one set of four frames that included a black car, a gray car, a blue car and, of all things, a dirty white pickup truck.
Needless to say, this could be a good feature if you could remove frames and it could learn, over time, some of the unique details of say the silver car that I could have identified and put a name to, if only I could have removed the three frames that had noticeable details that were different.  
Paul

[jguerdat]

You can remove them by using the 3 dot menu icon at the top right. You can then select specific cars to remove or to add to a known vehicle. Also do the same thing for Unrecognized.

[KachinaShadow]

@jguerdat 

Thanks...I'll try that.

Any idea if this will have the ability to learn, on an individual subscriber basis over a period of time, the differences in vehicles that it tries to identify as being one that you've already ID'd or is it just pretty basic such as just overall shape & color?

I realize that some of the differences are so subtle that it would be nearly impossible to be 100% accurate but, as I noted earlier, having four vehicles (three cars and a truck...all different colors) thrown in the same "bucket", seems that it still needs a little tweaking.

[jguerdat]

Subtle? Ha! FedEx and UPS always get mixed. I do hope there's learning going on to slowly stop this mix-up but time will tell. I don't monitor it that often so I may be slow-walking that supposed learning.

[KachinaShadow]

@jguerdat 

You forgot to throw Amazon in there too 😉

[KachinaShadow]

@jguerdat 

---

@jguerdat wrote:

Subtle? Ha! FedEx and UPS always get mixed. I do hope there's learning going on to slowly stop this mix-up but time will tell. I don't monitor it that often so I may be slow-walking that supposed learning.

---

Your comment got me thinking about this...so to make it easier I created one bucket titled Amazon/FedEx/UPS/USPS and threw the "recognized" images for all of them in there 😁

[jguerdat]

That certainly would make it easier. However, while I am currently using this feature, I don't pay much attention to it.