I thinks the new general V3.1 is working extremely well, it was able to identify the UPS truck as a vehicle at 41% accuracy (see left side of image). I do see sometimes false positives but is is rare.
Please let me know if you are seeing similar results
I think general v3.1 is working great for me. It has grabbed some catches that I had to look twice at. It caught a vehicle it said. I looked at the screen shot thinking it was mistaken. But on second look, it caught my neighbor backing out of their driveway between the "V" of two branches of a tree. I have never seen it Id anything that far away through the slit of a tree. I normally mask off my neighbors' driveway (I am only concerned about what is going past my house on the road) but did not think that was needed for that camera. Although it does seem to grab a couple of trees as a person. I would rather side on a slightly more mistaken id, than a complete miss. I know how to deal with those rare miss-ID objects with masking.
Basically, the general v3.1 is great for my cameras that I record that day to day traffic of people and vehicles that are driving around/past my property. If I want a more focused Id's on what is pulling into my driveway, I am still deciding on what model(s) to use. I have
Blue Iris for example, not alert me to a person detected in my drive when it's a child riding their bike up in my driveway (person+bicycle to cancel the alert). Now I am sure if someone now wants to rob my house they can use a bicycle during the day to get past my notifications but they would still be recorded though. It is just a lot less notification to distract me.
Anyway, again well-done gentlemen for your hard work and time making these models for us. I thank you!