search

roflcoopter / viseron

Self-hosted, local only NVR and AI Computer Vision software. With features such as object detection, motion detection, face recognition and more, it gives you the power to keep an eye on your home, office or any other place you want to monitor.
MIT License
1.76k stars 179 forks source link

Setting up streams from motion eye (Config help) #6

Closed JoeGandy closed 4 years ago

JoeGandy commented 4 years ago

Hey, Might be an obvious one Description of streaming url from motioneye:

Streaming URL provides MJPEG streaming. It can be used as a source for other applications that deal with video streams and know how to handle MJPEGs, or it can be used as the src attribute of an HTML tag.

I'm given the url in the format http://192.168.1.111:8082

Initial config for testing:

# See the README for the full list of configuration options.
cameras:
  - name: front_room
    host: 192.168.1.111
    port: 6082
    path: /
    width: 640
    height: 360
    fps: 15
    motion_detection:
      interval: 1
      trigger: false
    object_detection:
      interval: 1
      labels:
        - label: person
          confidence: 0.8

Then I get

[2020-09-04 08:20:01] [lib.camera.front_room ] [ERROR ] - Error starting decoder pipe! rtsp://None:None@192.168.1.111:6082/: Invalid data found when processing input

in the docker logs

Do you have any advice for setting this up?

roflcoopter commented 4 years ago

Ahh, looks like i have not accounted for cameras without authentication... Can you enable debug logging and paste me the full ffmpeg decoder command?

logging:
  level: debug
JoeGandy commented 4 years ago

Thanks for the quick reply!

[2020-09-04 08:45:51] [lib.camera.front_room ] [DEBUG ] - FFMPEG decoder command: ffmpeg -hide_banner -loglevel error -loglevel panic -avoid_negative_ts make_zero -fflags nobuffer -flags low_delay -strict experimental -fflags +genpts -stimeout 5000000 -use_wallclock_as_timestamps 1 -vsync 0 -rtsp_transport tcp -i rtsp://test:test@192.168.1.111:6082/ -f rawvideo -pix_fmt nv12 pipe:1

As a test I enabled authentication in motion eye, but still no luck

roflcoopter commented 4 years ago

I dont have an MJPEG camera to work with so i will need some help verifying the issue if thats okay. Can you run this for me and show me the output?

docker exec -it viseron ffmpeg -hide_banner -f mjpeg -i http://test:test@192.168.1.111:6082/ -f null /dev/null
JoeGandy commented 4 years ago

Sure no worries, seems to run fine

|||||||||||||||:~# docker exec -it viseron ffmpeg -hide_banner -f mjpeg -i http://test:test@192.168.1.111:6082/ -f null /dev/null Input #0, mjpeg, from 'http://test:test@192.168.1.111:6082/': Duration: N/A, bitrate: N/A Stream #0:0: Video: mjpeg, yuvj420p(pc, bt470bg/unknown/unknown), 800x480 [SAR 1:1 DAR 5:3], 25 tbr, 1200k tbn, 25 tbc Stream mapping: Stream #0:0 -> #0:0 (mjpeg (native) -> wrapped_avframe (native)) Press [q] to stop, [?] for help Output #0, null, to '/dev/null': Metadata: encoder : Lavf57.83.100 Stream #0:0: Video: wrapped_avframe, yuvj420p, 800x480 [SAR 1:1 DAR 5:3], q=2-31, 200 kb/s, 25 fps, 25 tbn, 25 tbc Metadata: encoder : Lavc57.107.100 wrapped_avframe frame= 442 fps= 20 q=-0.0 size=N/A time=00:00:17.68 bitrate=N/A speed=0.802x

roflcoopter commented 4 years ago

Great! Could you try without the -f mjpeg aswell?

docker exec -it viseron ffmpeg -hide_banner -i http://test:test@192.168.1.111:6082/ -f null /dev/null
JoeGandy commented 4 years ago

Assuming you mean -f mpjeg to be removed :)

||||||||:~# docker exec -it viseron ffmpeg -hide_banner -i http://test:test@192.168.1.111:6082/ -f null /dev/null Input #0, mpjpeg, from 'http://test:test@192.168.1.111:6082/': Duration: N/A, bitrate: N/A Stream #0:0: Video: mjpeg, yuvj420p(pc, bt470bg/unknown/unknown), 800x480 [SAR 1:1 DAR 5:3], 25 tbr, 25 tbn, 25 tbc Stream mapping: Stream #0:0 -> #0:0 (mjpeg (native) -> wrapped_avframe (native)) Press [q] to stop, [?] for help Output #0, null, to '/dev/null': Metadata: encoder : Lavf57.83.100 Stream #0:0: Video: wrapped_avframe, yuvj420p, 800x480 [SAR 1:1 DAR 5:3], q=2-31, 200 kb/s, 25 fps, 25 tbn, 25 tbc Metadata: encoder : Lavc57.107.100 wrapped_avframe frame= 192 fps= 20 q=-0.0 size=N/A time=00:00:07.68 bitrate=N/A speed=0.807x

roflcoopter commented 4 years ago

Yes thats what i meant... Thanks, I will do the necessarry fixes and push to a dev branch for you to try out. You are running the generic image yes?

JoeGandy commented 4 years ago

Yeah I just grabbed it from docker hub through the app page on unraid, so not quite sure how to test the dev branches!

roflcoopter commented 4 years ago

not sure how unraid works but i will push roflcoopter/viseron-generic:dev. Can you pick the tag or does it always just grab latest?

JoeGandy commented 4 years ago

I think i might be able to specify a tag, i'll give it a shot :)

I'll try roflcoopter/viseron:dev now

roflcoopter commented 4 years ago

its not up yet! ill let you know when it is.

JoeGandy commented 4 years ago

Okay I can confirm i can get to the different tags :) I tested with pulling roflcoopter/viseron:1.2.0 which worked, so give me a shout when :dev is good to go

Thanks! sorry was a bit keen!

roflcoopter commented 4 years ago

it is up now under roflcoopter/viseron:dev. you need to add stream_format: mjpeg to your camera configuration.

As i said earlier, i sadly dont have an mjpeg camera to test with so its untested to say the least

JoeGandy commented 4 years ago

It's looking to be getting better,still stimeout issue, but im not sure what this means?

[2020-09-04 09:38:20] [lib.camera.front_room ] [INFO ] - Resolution: 640x360 @ 15 FPS [2020-09-04 09:38:20] [lib.camera.front_room ] [DEBUG ] - FFMPEG decoder command: ffmpeg -hide_banner -loglevel error -loglevel panic -avoid_negative_ts make_zero -fflags nobuffer -flags low_delay -strict experimental -fflags +genpts -stimeout 5000000 -use_wallclock_as_timestamps 1 -vsync 0 -i http://test:test@192.168.1.111:6082/ -f rawvideo -pix_fmt nv12 pipe:1 [2020-09-04 09:38:20] [lib.camera.front_room ] [DEBUG ] - Starting capture process [2020-09-04 09:38:20] [lib.camera.front_room ] [DEBUG ] - Starting decoder thread [2020-09-04 09:38:20] [lib.recorder ] [DEBUG ] - Initializing ffmpeg recorder [2020-09-04 09:38:20] [lib.recorder ] [DEBUG ] - FFMPEG encoder command: ffmpeg -hide_banner -loglevel panic -f rawvideo -pix_fmt nv12 -s:v x -r -i pipe:0 -y [2020-09-04 09:38:20] [lib.nvr.front_room ] [DEBUG ] - NVR thread initialized [2020-09-04 09:38:20] [lib.nvr.front_room ] [DEBUG ] - Waiting for first frame [2020-09-04 09:38:20] [root ] [INFO ] - Initialization complete [2020-09-04 09:38:21] [lib.camera.front_room ] [ERROR ] - Error starting decoder pipe! Option stimeout not found. [2020-09-04 09:38:21] [lib.camera.front_room ] [ERROR ] - Error starting decoder pipe! Option stimeout not found.

Happy to keep debugging fyi :)

roflcoopter commented 4 years ago

Okay, guess the input parameters are not available for mjpeg streams. Can you run docker exec -it viseron ffmpeg -hide_banner -loglevel error -avoid_negative_ts make_zero -fflags nobuffer -flags low_delay -strict experimental -fflags +genpts -use_wallclock_as_timestamps 1 -vsync 0 -i http://test:test@192.168.1.111:6082/ -f null /dev/null

and see if there are any other offending flags? This way i dont have to keep pushing dev for you to find out some other flag is causing issues aswell

JoeGandy commented 4 years ago

Seems to be happy, dont get any ouput from the command

roflcoopter commented 4 years ago

great! You can set the input arguments for each camera through the config.yaml, like this

input_args:
  - -avoid_negative_ts
  - make_zero
  - -fflags
  - nobuffer
  - -flags
  - low_delay
  - -strict
  - experimental
  - -fflags
  - +genpts
  - -use_wallclock_as_timestamps
  - 1
  - -vsync
  - 0

try that and see if it works

JoeGandy commented 4 years ago

Hey, that's working now

[2020-09-04 10:29:08] [root ] [INFO ] - ------------------------------------------- [2020-09-04 10:29:08] [root ] [INFO ] - Initializing... [2020-09-04 10:29:08] [root ] [DEBUG ] - Starting cleanup scheduler [2020-09-04 10:29:08] [root ] [DEBUG ] - Running initial cleanup [2020-09-04 10:29:08] [lib.cleanup ] [DEBUG ] - Running cleanup [2020-09-04 10:29:08] [lib.mqtt ] [INFO ] - Initializing MQTT connection [2020-09-04 10:29:08] [lib.detector ] [INFO ] - Initializing detection thread [2020-09-04 10:29:08] [root ] [INFO ] - Initializing NVR threads [2020-09-04 10:29:08] [lib.nvr.front_room ] [DEBUG ] - Initializing NVR thread [2020-09-04 10:29:08] [lib.camera.front_room ] [DEBUG ] - Initializing ffmpeg RTSP pipe [2020-09-04 10:29:08] [lib.camera.front_room ] [INFO ] - Resolution: 640x360 @ 15 FPS [2020-09-04 10:29:08] [lib.camera.front_room ] [DEBUG ] - FFMPEG decoder command: ffmpeg -hide_banner -loglevel error -loglevel panic -avoid_negative_ts make_zero -fflags nobuffer -flags low_delay -strict experimental -fflags +genpts -use_wallclock_as_timestamps 1 -vsync 0 -i http://test:test@192.168.1.111:6082/ -f rawvideo -pix_fmt nv12 pipe:1 [2020-09-04 10:29:08] [lib.camera.front_room ] [DEBUG ] - Starting capture process [2020-09-04 10:29:08] [lib.camera.front_room ] [DEBUG ] - Starting decoder thread [2020-09-04 10:29:08] [lib.recorder ] [DEBUG ] - Initializing ffmpeg recorder [2020-09-04 10:29:08] [lib.recorder ] [DEBUG ] - FFMPEG encoder command: ffmpeg -hide_banner -loglevel panic -f rawvideo -pix_fmt nv12 -s:v x -r -i pipe:0 -y [2020-09-04 10:29:08] [lib.nvr.front_room ] [DEBUG ] - NVR thread initialized [2020-09-04 10:29:08] [lib.nvr.front_room ] [DEBUG ] - Waiting for first frame [2020-09-04 10:29:08] [root ] [INFO ] - Initialization complete [2020-09-04 10:29:09] [lib.camera.front_room ] [DEBUG ] - Running object detection at 1s interval [2020-09-04 10:29:09] [lib.camera.front_room ] [DEBUG ] - Running motion detection at 1s interval [2020-09-04 10:29:10] [lib.nvr.front_room ] [DEBUG ] - First frame received

I'm still not getting any events through mqtt other than lwt = alive but im sure thats just some config issues, im going to keep playing around with it :) Thanks a bunch so far!

JoeGandy commented 4 years ago

Hey apologies, just a couple more questions i hope,

I'm getting mqtt data through now, but im not getting any object detections. I get what id expect in the console:

[2020-09-04 13:02:28] [lib.camera.front_room ] [DEBUG ] - Running object detection at 1s interval [2020-09-04 13:02:28] [lib.camera.front_room ] [DEBUG ] - Running motion detection at 1s interval [2020-09-04 13:02:29] [lib.nvr.front_room ] [DEBUG ] - First frame received [2020-09-04 13:08:44] [lib.motion ] [DEBUG ] - Motion frames: 1, area: 175.5 [2020-09-04 13:08:44] [lib.motion ] [DEBUG ] - Motion frames: 2, area: 51912.0 [2020-09-04 13:08:44] [lib.nvr.front_room ] [DEBUG ] - Motion detected! Starting object detector [2020-09-04 13:08:45] [lib.motion ] [DEBUG ] - Motion frames: 3, area: 3781.5 [2020-09-04 13:08:45] [lib.motion ] [DEBUG ] - Motion frames: 4, area: 3052.5 [2020-09-04 13:08:46] [lib.motion ] [DEBUG ] - Motion frames: 5, area: 2691.0 [2020-09-04 13:08:46] [lib.motion ] [DEBUG ] - Motion frames: 6, area: 2639.0 [2020-09-04 13:08:46] [lib.motion ] [DEBUG ] - Motion frames: 7, area: 3486.5 [2020-09-04 13:08:47] [lib.motion ] [DEBUG ] - Motion frames: 8, area: 4360.0 [2020-09-04 13:08:47] [lib.motion ] [DEBUG ] - Motion frames: 9, area: 3429.0 [2020-09-04 13:08:48] [lib.motion ] [DEBUG ] - Motion frames: 10, area: 4331.0 [2020-09-04 13:08:48] [lib.motion ] [DEBUG ] - Motion frames: 11, area: 2369.5 [2020-09-04 13:08:49] [lib.motion ] [DEBUG ] - Motion frames: 12, area: 2193.5 [2020-09-04 13:08:49] [lib.motion ] [DEBUG ] - Motion frames: 13, area: 1583.5 [2020-09-04 13:08:50] [lib.motion ] [DEBUG ] - Motion frames: 14, area: 1245.0 [2020-09-04 13:08:50] [lib.motion ] [DEBUG ] - Motion frames: 15, area: 1043.5 [2020-09-04 13:08:51] [lib.motion ] [DEBUG ] - Motion frames: 16, area: 651.0 [2020-09-04 13:08:51] [lib.motion ] [DEBUG ] - Motion frames: 17, area: 472.0 [2020-09-04 13:08:51] [lib.motion ] [DEBUG ] - Motion frames: 18, area: 738.0 [2020-09-04 13:08:52] [lib.motion ] [DEBUG ] - Motion frames: 19, area: 4241.0 [2020-09-04 13:08:52] [lib.motion ] [DEBUG ] - Motion frames: 20, area: 2740.0 [2020-09-04 13:08:53] [lib.motion ] [DEBUG ] - Motion frames: 21, area: 2429.0 [2020-09-04 13:08:53] [lib.motion ] [DEBUG ] - Motion frames: 22, area: 1739.0 [2020-09-04 13:08:54] [lib.motion ] [DEBUG ] - Motion frames: 23, area: 1464.0 [2020-09-04 13:08:54] [lib.motion ] [DEBUG ] - Motion frames: 24, area: 2008.0 [2020-09-04 13:08:55] [lib.motion ] [DEBUG ] - Motion frames: 25, area: 3259.5 [2020-09-04 13:08:55] [lib.motion ] [DEBUG ] - Motion frames: 26, area: 3926.5 [2020-09-04 13:08:55] [lib.motion ] [DEBUG ] - Motion frames: 27, area: 4439.5 [2020-09-04 13:08:56] [lib.motion ] [DEBUG ] - Motion frames: 28, area: 2607.5 [2020-09-04 13:08:56] [lib.motion ] [DEBUG ] - Motion frames: 29, area: 3043.0 [2020-09-04 13:08:57] [lib.motion ] [DEBUG ] - Motion frames: 30, area: 4267.5 [2020-09-04 13:08:57] [lib.motion ] [DEBUG ] - Motion frames: 31, area: 2067.0 [2020-09-04 13:08:58] [lib.motion ] [DEBUG ] - Motion frames: 32, area: 1072.0 [2020-09-04 13:08:58] [lib.motion ] [DEBUG ] - Motion frames: 33, area: 744.5 [2020-09-04 13:08:59] [lib.motion ] [DEBUG ] - Motion frames: 34, area: 199.5 [2020-09-04 13:08:59] [lib.motion ] [DEBUG ] - Motion frames: 35, area: 97.5 [2020-09-04 13:09:00] [lib.motion ] [DEBUG ] - Motion frames: 36, area: 73.5 [2020-09-04 13:09:00] [lib.motion ] [DEBUG ] - Motion frames: 37, area: 336.0 [2020-09-04 13:09:00] [lib.motion ] [DEBUG ] - Motion frames: 38, area: 223.5 [2020-09-04 13:09:01] [lib.motion ] [DEBUG ] - Motion frames: 39, area: 181.0 [2020-09-04 13:09:01] [lib.motion ] [DEBUG ] - Motion frames: 40, area: 160.5 [2020-09-04 13:09:02] [lib.motion ] [DEBUG ] - Motion frames: 41, area: 249.5 [2020-09-04 13:09:02] [lib.motion ] [DEBUG ] - Motion frames: 42, area: 91.5 [2020-09-04 13:09:03] [lib.motion ] [DEBUG ] - Motion frames: 43, area: 52.5 [2020-09-04 13:09:03] [lib.motion ] [DEBUG ] - Motion frames: 44, area: 24.0 [2020-09-04 13:09:04] [lib.motion ] [DEBUG ] - Motion frames: 45, area: 32.5 [2020-09-04 13:09:04] [lib.motion ] [DEBUG ] - Motion frames: 46, area: 63.5 [2020-09-04 13:09:05] [lib.motion ] [DEBUG ] - Motion frames: 47, area: 46.5 [2020-09-04 13:09:05] [lib.motion ] [DEBUG ] - Motion frames: 48, area: 34.0 [2020-09-04 13:09:05] [lib.motion ] [DEBUG ] - Motion frames: 49, area: 28.5 [2020-09-04 13:09:06] [lib.motion ] [DEBUG ] - Motion has ended [2020-09-04 13:09:06] [lib.nvr.front_room ] [DEBUG ] - Not recording, pausing object detector

But I don't get any recordings saved in the /recordings folder, or any detections in the mqtt data/home assistant sensor. Also the memory usage seems to consistently climb until it uses up everything i've got, I have lookback set to 0, that could be a problem?

Config below (its deliberately really sensitive for testing! :) ) 

cameras:
  - name: front_room
    host: 192.168.1.111
    port: 6082
    path: /
    width: 640
    height: 360
    username: test
    password: test
    stream_format: mjpeg
    fps: 15
    input_args:
      - "-avoid_negative_ts"
      - "make_zero"
      - "-fflags"
      - "nobuffer"
      - "-flags"
      - "low_delay"
      - "-strict"
      - "experimental"
      - "-fflags"
      - "+genpts"
      - "-use_wallclock_as_timestamps"
      - "1"
      - "-vsync"
      - "0"

motion_detection:
  interval: 1
  trigger: true
  area: 10
  frames: 2

object_detection:
  interval: 1
  labels:
    - label: person
      confidence: 0.1
    - label: cat
      confidence: 0.1

recorder:
  lookback: 0
  timeout: 10
  retain: 7
  folder: /recordings

# MQTT is optional
mqtt:
  broker: <removed>
  port: <removed>
  username: <removed>
  password: <removed>

logging:
  level: debug
JoeGandy commented 4 years ago

FYI, lookback set to 1 fixed the memory issue atleast! I'm an idiot! still not getting detections :)

roflcoopter commented 4 years ago

Okay i will look at the memory issue when set to 0. Have you tried walking in front of the camera?

I am working on making it easier to debug the detections.

JoeGandy commented 4 years ago

Okay cool, Yeah i think i the lookback is 0 its prob just storing the feed forever in memory. Yep as you can see i get the motion events just fine! Just nothing to debug the object detection, is there a log anywhere for the actual detectors?

Happy to close this issue now, as its getting a bit off topic, and if you're already working on it i can't ask for much more!

roflcoopter commented 4 years ago

Yes its allocation a python queue and when the size is zero it means infinite. Didnt account for this...

If you set trigger: false under motion detected, are there still no detections?

JoeGandy commented 4 years ago

Nope still nothing, feels like object detection isnt even running, do you have any commands I could try to confirm its even setup right?

roflcoopter commented 4 years ago

There should be errors if there is something wrong. What i need to do is publish some debug information to MQTT for you to inspect, but i wont have time to do that today. I will try to squeeze it in during the weekend. Sorry for the trouble!

JoeGandy commented 4 years ago

Haha no worries! take as long as you need, appreciate the work you've done on this so far, ill keep trying some stuff over the weekend too

JoeGandy commented 4 years ago

Hey, Did you have any luck with this @roflcoopter If i could get some way to debug the actual object dection that'd be awesome.

I can have a look at the unraid template too for you :)

roflcoopter commented 4 years ago

Sorry for taking so long, I am working on zones atm which needed quite a big refactor of the codebase. When this is done images will be sent to MQTT with all objects drawn together with configured zones.

I could push a few debug log statements tomorrow morning so you can see if the detection is running at all.

roflcoopter commented 4 years ago

I just published the first beta for zones.

I also added debug logging of the result for each processed frame by the detector. Here is an example

[2020-09-08 10:00:16] [lib.motion              ] [DEBUG   ] - Motion frames: 1, area: 768.0
[2020-09-08 10:00:16] [lib.nvr.uteplats        ] [DEBUG   ] - Motion detected! Starting object detector
[2020-09-08 10:00:16] [lib.nvr.uteplats        ] [DEBUG   ] - Objects: [{'label': 'pottedplant', 'confidence': 0.224, 'height': 0.062, 'width': 0.065, 'relative_x1': 0.673, 'relative_y1': 0.368, 'relative_x2': 0.738, 'relative_y2': 0.43}]
[2020-09-08 10:00:17] [lib.motion              ] [DEBUG   ] - Motion frames: 2, area: 2520.5
[2020-09-08 10:00:17] [lib.nvr.uteplats        ] [DEBUG   ] - Objects: [{'label': 'pottedplant', 'confidence': 0.296, 'height': 0.062, 'width': 0.065, 'relative_x1': 0.675, 'relative_y1': 0.368, 'relative_x2': 0.74, 'relative_y2': 0.43}]
[2020-09-08 10:00:18] [lib.motion              ] [DEBUG   ] - Motion frames: 3, area: 293.5
[2020-09-08 10:00:18] [lib.nvr.uteplats        ] [DEBUG   ] - Objects: [{'label': 'pottedplant', 'confidence': 0.233, 'height': 0.06, 'width': 0.067, 'relative_x1': 0.671, 'relative_y1': 0.37, 'relative_x2': 0.738, 'relative_y2': 0.43}]
[2020-09-08 10:00:19] [lib.motion              ] [DEBUG   ] - Motion frames: 4, area: 996.5
[2020-09-08 10:00:19] [lib.nvr.uteplats        ] [DEBUG   ] - Objects: []
[2020-09-08 10:00:20] [lib.motion              ] [DEBUG   ] - Motion frames: 5, area: 1542.5
[2020-09-08 10:00:20] [lib.nvr.uteplats        ] [DEBUG   ] - Objects: [{'label': 'pottedplant', 'confidence': 0.203, 'height': 0.062, 'width': 0.067, 'relative_x1': 0.675, 'relative_y1': 0.368, 'relative_x2': 0.743, 'relative_y2': 0.43}]
[2020-09-08 10:00:21] [lib.motion              ] [DEBUG   ] - Motion frames: 6, area: 1498.0
[2020-09-08 10:00:21] [lib.nvr.uteplats        ] [DEBUG   ] - Objects: []
[2020-09-08 10:00:22] [lib.motion              ] [DEBUG   ] - Motion frames: 7, area: 1337.5
[2020-09-08 10:00:22] [lib.nvr.uteplats        ] [DEBUG   ] - Objects: []

I am closing this issue as its gone quite offtopic. Feel free to open a new issue if you still are having issues with the detection