Blue Iris motion detection suddenly not working at night!

thanks for the reply, I think i did it but i would like confirmation. I did add my geolocation. I set the slider to the exact time of the sunrise on my night profile and i did get the checkbox. When i hit ok (box closes) and then open the schedule up to view it, the box is unchecked. Is this the way it is supposed to work? I will attached 2 pics. thanks
View attachment 16533View attachment 16534
You should be ok...
 
I just installed my Hikvision outside and am having the same issue. What settings do you need to assist please?
 
thanks to this forum i learned i needed to create a second schedule and i called it night mode. Once i created the night mode I went into one of my cams and cranked up the sensitivity at sunset - sunrise. It works great for me now. I even detect wandering cats and dogs in my front yard at all hours in the night.
 
As a suggestion....
For testing.... Copy a camera into 2 - 4, copies. Each copy having a slightly different set of settings. Walk around the area in question to test triggering/alerting. This will reduce your testing time, since your testing the SAME camera with 4 different configurations simultaneously!

I've got a few 2000+ lumen led 180d spotlights around my house - dusk to dawn with motion detection. This has been enough to scare away the trouble makers... so far. You should see the look on their faces when the spotlights go to full brightness! lol
 
When you say post your settings, is there an easy way to export my settings? If not what settings are you looking for?
sceenshot all you motion settings..if you are using more than one profile you need to do this for each profile...
 
Hi..
I know this is a very old thread…but I am on the latest version of bl using Deepstack which has been working great, except I notice that some nights the motion doesn’t work at all. My cameras have color night vision and when I set up Deepstack and tested it even at night, it worked great and motion alerts were almost instant.

The way I noticed it doesn’t work at night is because my son waits for the school bus every morning right outside the door and the camera is about 15 feet away. Now that is dark at 6:30 am, I don’t get motion when he’s standing there vs when it was daytime.

During the day motion works flawlessly.
I’ve been reading about creating a schedule for night, but Im not sure what setting I need to change for night time.
 
@macster2075 - since you are now running DeepStack, most of us have found that using the Dark custom model results in better recognition of an object compared to the default DeepStack model. You can find threads here and in the BI help file on how to set that up.

Also, these are field of view dependent and requires the camera to be set up properly (not on default/auto settings) in order to have the highest chance of object detection. If there is not enough contrast or the object is on the fringes of the field of view, it can be problematic at night.
 
@macster2075 - since you are now running DeepStack, most of us have found that using the Dark custom model results in better recognition of an object compared to the default DeepStack model. You can find threads here and in the BI help file on how to set that up.

Also, these are field of view dependent and requires the camera to be set up properly (not on default/auto settings) in order to have the highest chance of object detection. If there is not enough contrast or the object is on the fringes of the field of view, it can be problematic at night.
Oh ok.. didn’t know there was a dark custom option..I am fairly new to bl and deepstack…I will def look into that. Thanks!
 
It is simple to do. Create a Mymodels folder and past the dark.pt model into it and then tell DS to use custom models and then in object detection add the wording for the custom model (People, Car). The dark model starts with a capital letter and person is People in that model.

Also, some have trained a model to their particular camera and field of view. That will result in the best object detection.
 
Now that Im off work I have been searching for the Dark custom model for Deepstack, but in all honesty.. I don't know what Im really looking for or were do I even make that change in BL haha.
 
yup.. I found that...but I still don't really understand what I need to do...all those codes and things I don't really understand.
 
I was hoping to maybe find a video on how to do this... I guess is very simple for someone who's done it before, but for a simpleton like me...it's in another language lol
 
I believe I have found the video of the same guy in Youtube, but there's no talking...just him typing a bunch of stuff not explaining what he's doing.. I learned nothing from it.
 
Yeah, the documentation is lacking.

Go to this site:


And download the dark.pt model

Then go to your DeepStack folder and create a MyModels folder and past dark.pt into it.

Then go to BI under the global setup and tell it to use custom models.

Then under the cameras you want to use the dark model, go into each camera and type in the objects you want detected.

After you set it up, it is good to reboot your computer to make sure it starts fresh and everything works.
 
Thanks.. I'll give that a shot... so in the AI section in BL I already have "person,car,truck,bus,bicycle,cat,dog" -
So that means I don't have to add anything else right?

Also, I've was wondering the objects on the list of things to be detected in DS.. toothbrush?.. really? lol
Why would anyone want to detect a toothbrush?

So using a method like this, I can train my own DS to detect anything I want right? - but I still have to create that rule first?
 
Last edited:
For the dark model, you need to add onto that with the correct word for the dark model. In most instances it is just a capital first letter.

  • Bicycle
  • Boat
  • Bottle
  • Bus
  • Chair
  • Car
  • Cat
  • Cup
  • Dog
  • Motorbike
  • People
  • Table
But person is People and not Person for the dark model.

So if you wanted to get a human and a dog - you would put into the confirm box - dog,Dog,person,People

To train a model, you basically create your own model using the pictures from your field of view. Someone here said it take an hour or so to create a model.