ENVs I changed:
CUDA_PKG_VERSION = 11-1=11.1.1
CUDA_VERSION = 11.1.1
CUDNN_VERSION = 8.0.5.39
NVIDIA_REQUIRE_CUDA = cuda>=11.1 brand=tesla,driver>=396,driver<397 brand=tesla,driver>=410,driver<411 brand=tesla,driver>=418,driver<419 brand=tesla,driver>=440,driver<441
Edit:
… and in LABELS:
com.nvidia.cudnn.version = 8.0.5.39
You might be in luck then! 3.5 is still supported in CUDA 11.
It takes a few hours to rebuild everything, i will let you know when its done and you can test it out
Great! Thank you! 
The builds are now complete, can you pull the dev tag and test if it works for you?
Hi,
First of all, I would like to congratulate you for this awesome project.
I intented to integrate it into home assistant but i met some issues with MQTT integration
I use the Mosquitto Broker addon and my viseron config file looks like this
MQTT is optional
mqtt:
broker: core-mosquitto
port: 1883
username : user
password: pass
Viseron logs are the following
[2021-06-05 16:52:44] [lib.camera.front_door ] [DEBUG ] - Getting stream information for rtsp:/[email protected]:554//h264Preview_01_main,
[2021-06-05 16:52:44] [lib.camera.front_door ] [DEBUG ] - Initializing camera Front Door,
[2021-06-05 16:52:44] [lib.nvr.front_door ] [DEBUG ] - Initializing NVR thread,
[2021-06-05 16:52:44] [lib.mqtt ] [INFO ] - Initializing MQTT connection,
[2021-06-05 16:52:44] [root ] [INFO ] - Initializing NVR threads,
[2021-06-05 16:52:44] [lib.detector ] [DEBUG ] - Initializing object detector darknet,
[2021-06-05 16:52:44] [lib.detector ] [DEBUG ] - OpenCL activated,
[2021-06-05 16:52:44] [lib.detector ] [DEBUG ] - Object detector initialized
No errors seem to appear but the elements from my camera are not discovered by MQTT into home assistant
Any suggestion ?
Thanks for the kind words!
Have you changed the MQTT discovery prefix in Home Assistant?
Could you post your full config wrapped in triple backticks? (```)
Thanks for having replied
Why is it necessary to change the MQTT discovery prefix ? Does Viseron use a specific prefix for MQTT ?
What I do not understand is why I cannot see in Mosquitto Broker logs the IP address of Viseron (172.17.0.3)
The IP address of the MQTT broker is the same as my server (all containers are on the same sever) ?
I have tried to follow the MQTT discovery tutorial, but it unfortunately did not work for me (the test is stuck)
Viseron config.yaml
# See the README for the full list of configuration options.
cameras:
- name: Front Door
mqtt_name: viseron_front_door
host: !secret camera_ip
port: 554
username: !secret username
password: !secret password
path: //h264Preview_01_main
fps: 30
width: 2560
height: 1920
stream_format: rtsp
publish_image: true
motion_detection:
interval: 1
trigger_detector: false
# timeout: true
# max_timeout: 30
# width: 300
# height: 300
# area: 0.1
# frames: 3
object_detection:
interval: 1
labels:
- label: person
confidence: 0.8
- label: cat
confidence: 0.6
recorder:
lookback: 60
timeout: 120
retain: 365
folder: /recordings
thumbnail:
save_to_disk: true
send_to_mqtt: true
# MQTT is optional
mqtt:
broker: 192.168.10.2
port: 1883
username: XXX
password: XXX
logging:
level: debug
Viseron logs:
[2021-06-06 14:43:16] [root ] [INFO ] - -------------------------------------------,
[2021-06-06 14:43:16] [root ] [INFO ] - Initializing...,
[2021-06-06 14:43:16] [root ] [DEBUG ] - Starting cleanup scheduler,
[2021-06-06 14:43:16] [root ] [DEBUG ] - Running initial cleanup,
[2021-06-06 14:43:16] [lib.cleanup ] [DEBUG ] - Running cleanup,
[2021-06-06 14:43:16] [lib.cleanup ] [DEBUG ] - Items in /recordings/2021-06-04: 1,
[2021-06-06 14:43:16] [lib.cleanup ] [DEBUG ] - Items in /recordings/2021-06-05: 1,
[2021-06-06 14:43:16] [lib.cleanup ] [DEBUG ] - Items in /recordings/2021-05-31: 1,
[2021-06-06 14:43:16] [lib.cleanup ] [DEBUG ] - Items in /recordings/2021-06-06: 1,
[2021-06-06 14:43:16] [lib.mqtt ] [INFO ] - Initializing MQTT connection,
[2021-06-06 14:43:16] [lib.detector ] [DEBUG ] - Initializing object detector darknet,
[2021-06-06 14:43:16] [lib.detector ] [DEBUG ] - OpenCL activated,
[2021-06-06 14:43:16] [lib.detector ] [DEBUG ] - Object detector initialized,
[2021-06-06 14:43:16] [root ] [INFO ] - Initializing NVR threads,
[2021-06-06 14:43:16] [lib.nvr.front_door ] [DEBUG ] - Initializing NVR thread,
[2021-06-06 14:43:16] [lib.camera.front_door ] [DEBUG ] - Initializing camera Front Door,
[2021-06-06 14:43:16] [lib.camera.front_door ] [DEBUG ] - Getting stream information for rtsp://xxxx:[email protected]:554//h264Preview_01_main,
MQTT broker logs:
[s6-init] making user provided files available at /var/run/s6/etc...exited 0.,
[s6-init] ensuring user provided files have correct perms...exited 0.,
[fix-attrs.d] applying ownership & permissions fixes...,
[fix-attrs.d] done.,
[cont-init.d] executing container initialization scripts...,
[cont-init.d] mosquitto.sh: executing... ,
[13:12:56] INFO: SSL is not enabled,
[cont-init.d] mosquitto.sh: exited 0.,
[cont-init.d] nginx.sh: executing... ,
[cont-init.d] nginx.sh: exited 0.,
[cont-init.d] done.,
[services.d] starting services,
[services.d] done.,
[13:12:57] INFO: Starting NGINX for authentication handling...,
[13:12:57] INFO: Starting mosquitto MQTT broker...,
1622977977: mosquitto version 1.6.12 starting,
1622977977: |-- *** auth-plug: startup,
[13:12:57] INFO: Successfully send discovery information to Home Assistant.,
[13:12:57] INFO: Successfully send service information to the Supervisor.,
1622977977: Config loaded from /etc/mosquitto/mosquitto.conf.,
1622977977: Loading plugin: /usr/share/mosquitto/auth-plug.so,
1622977977: ├── Username/password checking enabled.,
1622977977: ├── TLS-PSK checking enabled.,
1622977977: └── Extended authentication not enabled.,
1622977977: Opening ipv4 listen socket on port 1883.,
1622977977: Opening ipv6 listen socket on port 1883.,
1622977977: Opening websockets listen socket on port 1884.,
1622977977: Warning: Mosquitto should not be run as root/administrator.,
1622977977: mosquitto version 1.6.12 running,
1622977977: New connection from 127.0.0.1 on port 1883.,
1622977977: Socket error on client <unknown>, disconnecting.,
1622977983: New connection from 172.30.33.2 on port 1883.,
1622977983: New client connected from 172.30.33.2 as mosq-KhN02P1FYxJ093adpz (p2, c1, k60, u'addons').,
1622977983: Client mosq-KhN02P1FYxJ093adpz disconnected.,
1622977984: New connection from 172.30.33.2 on port 1883.,
1622977984: New client connected from 172.30.33.2 as mosq-3vb54eT5JHbOAbMU9J (p2, c1, k60, u'addons').,
1622977984: Client mosq-3vb54eT5JHbOAbMU9J disconnected.,
1622977984: New connection from 172.30.33.2 on port 1883.,
1622977984: New bridge connected from 172.30.33.2 as zwave (p2, c1, k60, u'addons').,
1622977990: New connection from 172.30.32.1 on port 1883.,
{"result": "ok", "data": {}}1622977990: New client connected from 172.30.32.1 as 4kFVBcF1f9c2xzPbl6UESR (p2, c1, k60, u'Matthieu').,
1622978016: New connection from 172.30.33.3 on port 1883.,
1622978016: New client connected from 172.30.33.3 as mqttjs_9829a5e9 (p2, c1, k60, u'addons').,
1622979778: Saving in-memory database to /data/mosquitto.db.,
1622981579: Saving in-memory database to /data/mosquitto.db.,
1622983299: New connection from 172.30.33.1 on port 1883.,
1622983299: New client connected from 172.30.33.1 as mosq-cRVebsMRVAl91oTcja (p2, c1, k60).,
1622983299: Client mosq-cRVebsMRVAl91oTcja disconnected.,
1622983363: New connection from 172.30.32.1 on port 1883.,
1622983363: New client connected from 172.30.32.1 as mosq-pdG4WLHei7q3cxsamc (p2, c1, k60).,
Il really suspect a problem of connexion between viseron and MQTT.
Indeed, when I put MQTT elements into config.yaml, detection do not work anymore (even if logs seem to be ok and no connexion errors with MQTT are mentionned)
Unfortunately it does not work.
Please find attached logs.
I apologize for the inconveniences, but I have found my problem
Actually, to solve it I had to :
Hope it will help someone else
Anyway, thank you for your help
could you post your config?
Great you solved it!
Regarding the codec not being detected, what docker image are you pulling? Can you show your docker run command?
cameras:
- name: raspicam
host: 192.168.1.14
port: 8555
rtsp_transport: tcp
username: user
password: password
path: /unicast
width: 1280
height: 720
fps: 10
Does it work if you manually specify the codec? Use h264_cuvid for h264 or hevc_cuvid for h265.
cameras:
- name: raspicam
host: 192.168.1.14
port: 8555
rtsp_transport: tcp
username: user
password: password
path: /unicast
width: 1280
height: 720
fps: 10
codec: h264_cuvid
Also looking at the logs it seems you hit a bug when running without MQTT, ill fix that tomorrow morning
The image is
roflcoopter/viseron-vaapi:latest from 2021-05-21
sha256:d61e41077b4c853c636a4cd3d8685a7b071717deab47a493dc96513ea27003aa
The running command is
docker run --name=viseron --hostname=b8dd4c3d9777 --mac-address=02:42:ac:11:00:0 3 --env=PATH=/root/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin --env=DEBIAN_FRONTEND=noninteractive --env=opencv=master --env=VISERON_CUDA_SUPPORTED=false --env=VISERON_VAAPI_SUPPORTED=true --env=VISERON_OPENCL_SUPPORTED=true --env=VISERON_RASPBERRYPI3=false --volume=/nvr:/recordings:rw --volume=/nvr/config:/config:rw --volume=/etc/localtime:/etc/localtime:ro --volume=/recordi ngs --workdir=/src/viseron --restart=always --device /dev/dri:/dev/dri --log-driver=journald --detach=true roflcoopter/viseron-vaapi:latest viseron.py
Besides, fyi, h264_cuvid is not recognized (that is why I have put h264)
[2021-06-06 22:24:26] [lib.camera.front_door ] [ERROR ] - Error starting decoder command! Unknown decoder 'h264_cuvid'
I tried both codecs and the container is terminating itself. I also specified these codec names for hwaccel_args argument and I can see only errors. In both cases I can see OpenCL and CUDA cannot be used messages.
I just realized you are pulling the incorrect image. If you want to use your GPU you need to pull roflcoopter/amd64-cuda-viseron:dev
Sorry for being unclear!
I suspect it might show that CUDA is not supported anyway but that should only be an issue with the automatic detection
You should change the image you are using, roflcoopter/viseron-vaapi:latest is old and wont be updated any longer.
Use roflcoopter/viseron:latest in the future!
No problem. I tried roflcoopter/amd64-cuda-viseron:dev yesterday (before viseron:dev) and there was info that CUDA could not be used. So I did not check mem status in NVTOP and switched immediately to viseron:dev. I will try it again and also with codec argument in the config. I will check nvtop as well. Thank you.
Edited:
Please find attached link to logs information: jarekmor@ubuntu_cuda:~$ docker run --rm -v /home/jarekmor/viseron/recordings:/re - Pastebin.com
There is indication in the nvtop that Nvidia is performing - 25% (GPU Mem%) - when movement is detected but still “CUDA can not be used” message.
I run it on basic config - no additional arguments. With additional arguments (codec, hwaccel_args) it does not work or prints errors.
There is also few minutes delay from movement detection to files recording time.
Also, as the first movie is recorded then after 1-2 minutes later the pictures with recognised objects are saved.
You can also find in the logs at the end that that Viseron is “Restarting frame pipe”.
After that error there is no recording at all.
Can you add debug logging to your config?
logging:
level: debug