Object Detection: Frigate vs Doods vs BlueIris vs Deepstack - with Coral

Third party integrations
1

Hello,

I was wondering if anyone has tried all of these: Frigate vs Doods vs BlueIris vs Deepstack with Google Coral for object detection and could give us a summary of Pros / Cons for each of them ?

2

By the way, do you have a current source of coral usb sticks?

3

I ordered (and paid for) one in November last year at the mouser webshop. Once a month I get an update: being produced. Scheduled at the moment to be received at the end of march.

4

I only use doods(June 21) and frigate (current) with coral

Doods was nice but UI and feature was lacking. Using this I did not feel confident to drop my motioneye as it was not full NVR(24/7 recording and ui for playback)

Frigate has great UI and still active development and improvement. I use this 1 week and stop use of a NVR. Only con is camera does not show in full resolution for live video. Playback is full resolution and has event showing along timeline. Config files are used for setup, feels like HA setup in beginning.

Both doods and frigate detection is decent but not perfect. Both allow use of your own model so this would help I expect.

I use blueiris and windows requirement is pain. I do not use VMs. I wasn’t crazy about it’s setup and use either. I did not use with coral but at the time.

5

In my opinion blue iris doesn’t really fit with the rest of them as it is a full blown NVR with a ton of customization. It also uses deepstack for its smart detection.

I use blue iris for continuous recording and backup recording of activity.

My opinions on the others

Doods:

  • I like that is continuously runs its detection on a given camera stream.
  • setup was simple enough following instructions + youtube videos
  • Ultimately I stopped using in favor of deepstack because after running them side by side I was getting more false alerts with doods.
  • No UI

Deepstack:

  • I don’t like that it needs to be triggered by something. I tried using mqtt motion messages from blue iris but this was to slow. Currently running an automation for each camera to trigger it every 2 seconds which does a pretty good job but is a rather clunky way of doing things.
  • setup was about the same as doods
  • Object detection is fairly good especially with some tweaking of confidence levels.
  • No UI to make adding masks easier, thus I don’t use them which leads to more false positives
  • This is what I am currently using and am mostly happy with it.

Frigate

  • UI is great
  • continuous detection is great
  • Detections seems good but as I don’t have a coral yet I shut it down because this plus deepstack was really hammering my CPU
  • I suspect once I get a coral this will be my way forward given the ease of adding face detection but until I get a coral its to CPU intensive to use as I add more cameras.
3 Likes
6

Thanks @Anemoi, great summary.

Regarding Coral:

  • Do you know of any benchmarks of the scale of the performance benefits?
  • Which systems is it compatible with?
7
1 Like
8

No problem I would really like to get my camera system consolidated and the software’s organized so that was pretty much putting my scattered thoughts on paper.

Now if I could just get ahold of a Coral

9

USB corals going for >$250 on eBay :sob:

10

I refuse to pay that.

I managed to back order one about 2 months ago but I don’t think there is even an estimated ship date.

I had a chance to buy on for $100 that I probably should have taken.

11

Thanks @tmjpugh!

Those numbers are definitely promising, but I was thinking of real-world improvements on the specific object detection stacks in this thread.

Oh well, I guess when I finally get my hands on one (!), I’ll try to compare.

1 Like
12

I backordered a Coral USB accelerator from here https://www.newark.com/coral/g950-01456-01/usb-accelerator-raspberry-pi-rohs/dp/26AJ4884

$108 + shipping and tax - should be available again in 2 months…

13

Good luck with that…

14

I talked with a friend from Google today and he told me the Coral program was cancelled because it did not yield enough profit, and the engineer who was lead on these chips left google a few months ago for another company.

So I don’t think we’ll see any more of these parts. If this is correct (my friend is very knowledge about this program), then what is the future of Frigate?

1 Like
15

Expensive overpriced scalpers on ebay.

1 Like
16

That’s rather unfortunate. :frowning:

Any chance Frigate might support CUDA?

Are there mobile / power efficient / cost effective Nvidia GPUs with compute capability >= 3.5?

(Secret option: use Tensorflow JS with a larger variety of GPUs.??)

17

Is asus going to make them on their own ?

https://iot.asus.com/products/AI-accelerator/

2 Likes
18

Wow has anyone been able to confirm that the Google Coral devices are being discontinued??

I just received a notification that my backorder is on the way and looks like it should arrive in a few weeks.

19

Asus for manufacturing and distribution in the future it seems

EDIT
Coral AI is needed part of their future business plans so I believe the will keep it alive in a usable form for general consumer

1 Like
20

Glad to see its not dying