viseron
viseron copied to clipboard
roflcoopter
Issue with vf transpose in Recorder
Since my camera is mounted sideways I have to rotate the image. After a lot of trail and an error I found out how to do with with this config:
recorder:
lookback: 3
timeout: 3
retain: 7
filter_args:
- '-vf'
- "transpose=2"
Rest of config
...
object_detection:
type: darknet
labels:
- label: person
confidence: 0.9
- label: truck
confidence: 0.9
- label: car
confidence: 0.9
mqtt:
# redacted
logging:
level: debug
cameras:
- name: Ute
mqtt_name: ute
host: 192.168.0.x
port: 554
username: admin
password: "..."
path: '/h264'
height: 2592
width: 1944
fps: 5
audio_codec: aac
publish_image: true
filter_args:
- '-vf'
- "transpose=2"
object_detection:
interval: .2
mask:
- points:
- x: 7
y: 515
- x: 152
y: 453
- x: 101
y: 403
- x: 30
y: 445
substream:
path: '/h264cif'
port: 554
height: 704
width: 576
fps: 5
filter_args:
- '-vf'
- "transpose=2"
static_mjpeg_streams:
small:
width: 100
height: 100
But after the motion and object detection has run I get this error and no mp4-file.
file 'file:/segments/Ute/20211123140136.mp4'
file 'file:/segments/Ute/20211123140148.mp4'
file 'file:/segments/Ute/20211123140200.mp4'
file 'file:/segments/Ute/20211123140212.mp4'
outpoint 5
Filtergraph 'transpose=2' was defined for video output stream 0:0 but codec copy was selected.
Filtering and streamcopy cannot be used together.
[2021-11-23 14:02:46] [ERROR ] [viseron.recorder.ute ] - Failed to concatenate segments: Command '['ffmpeg', '-hide_banne\
r', '-loglevel', 'error', '-y',
'-protocol_whitelist', 'file,pipe', '-f', 'concat', '-safe', '0', '-i', '-', '-c:v', 'copy', '-c:a', 'copy', '-vf', 'transpose\
=2', '-movflags', '+faststart',
'/recordings/2021-11-23/Ute/11:49:43.mp4']' returned non-zero exit status 1.
If I remove the -vf transpose=2 from the recorder it works, but then my recording is stored with the original format (sideways)
You can probably solve this by adding a codec like h264 to your recorder config.
Ok, let me try that. But also now that I have more than one camera, I really would need to specify this per camera level and not the global recorder:
That is coming in the next release. I am working on a huge rewrite of Viseron entirely, thats why there has not been any releases in a while.
It focuses on decoupling things, making it super easy to implement new functionality using components similar to Home Assistant if you are familiar with that
That's awesome to hear. I'm very familiar with Home Assistant, so that seems like a good approach.
You might also consider to hot reload the config for faster change time.
And while we talk about feature requests, I really would love to control the states from MQTT/Home Assistant. At the moment, since my camera is overlooking the driveway where I park my cars, it never stops recording. So for me - since I have the rule engine/state machine/automation engine in HA, I could automate start/stop based on motion sensors or when the person is not detected anymore.
You can kinda do this already by using require_motion on the tracked label.
By using that, stationary objects, like parked cars, wont keep the recorder going forever.
You can also use the MQTT switch enity in Home Assistant to completely shut off a camera, bascially disarming it until you turn it back on again
So back to the first problem. Did what you said with setting the codec, but now I getting the container killed. See logs:
[2021-11-26 19:33:41] [DEBUG ] [viseron.nvr.ute ] - Not recording, pausing object detector
[2021-11-26 19:33:41] [DEBUG ] [viseron.nvr.ute.object ] - Objects: [{'label': 'car', 'confidence': 0.991, 'rel_width': 0.457, 'rel_height': 0.375, 'rel_x1': 0.173, 'rel_y1': 0.541, 'rel_x2': 0.63, 'rel_y2': 0.916}, {'label': 'car', 'confidence': 0.982, 'rel_width': 0.314, 'rel_height': 0.162, 'rel_x1': 0.505, 'rel_y1': 0.507, 'rel_x2': 0.819, 'rel_y2': 0.669}]
[2021-11-26 19:33:45] [DEBUG ] [viseron.recorder.ute ] - /segments/Ute/20211126193340.mp4 is locked. Trying again in 1 second, message repeated 5 times
[2021-11-26 19:33:47] [DEBUG ] [viseron.recorder.ute ] - Segment information: {'20211126193328.mp4': {'start_time': 1637951608.0, 'end_time': 1637951614.081}, '20211126193252.mp4': {'start_time': 1637951572.0, 'end_time': 1637951578.002}, '20211126193304.mp4': {'start_time': 1637951584.0, 'end_time': 1637951590.207}, '20211126193258.mp4': {'start_time': 1637951578.0, 'end_time': 1637951584.007}, '20211126193340.mp4': {'start_time': 1637951620.0, 'end_time': 1637951626.071}, '20211126193316.mp4': {'start_time': 1637951596.0, 'end_time': 1637951602.027}, '20211126193334.mp4': {'start_time': 1637951614.0, 'end_time': 1637951620.153}, '20211126193310.mp4': {'start_time': 1637951590.0, 'end_time': 1637951596.089}, '20211126193322.mp4': {'start_time': 1637951602.0, 'end_time': 1637951608.089}}
[2021-11-26 19:33:47] [DEBUG ] [viseron.recorder.ute ] - Start event: 1637951582, segment: 20211126193258.mp4
[2021-11-26 19:33:47] [DEBUG ] [viseron.recorder.ute ] - End event: 1637951620, segment: 20211126193340.mp4
[2021-11-26 19:33:47] [DEBUG ] [viseron.recorder.ute ] - Concatenation command: ['ffmpeg', '-hide_banner', '-loglevel', 'error', '-y', '-protocol_whitelist', 'file,pipe', '-f', 'concat', '-safe', '0', '-i', '-', '-c:v', 'h264', '-c:a', 'copy', '-vf', 'transpose=2', '-movflags', '+faststart', '/recordings/2021-11-26/Ute/19:33:17.mp4']
[2021-11-26 19:33:47] [DEBUG ] [viseron.recorder.ute ] - Segment script:
file 'file:/segments/Ute/20211126193258.mp4'
inpoint 4
file 'file:/segments/Ute/20211126193304.mp4'
file 'file:/segments/Ute/20211126193310.mp4'
file 'file:/segments/Ute/20211126193316.mp4'
file 'file:/segments/Ute/20211126193322.mp4'
file 'file:/segments/Ute/20211126193328.mp4'
file 'file:/segments/Ute/20211126193334.mp4'
file 'file:/segments/Ute/20211126193340.mp4'
outpoint 0
./run: line 10: 343 Killed justc-envdir /var/run/environment exec s6-setuidgid abc python3 -u -m viseron
[cont-finish.d] executing container finish scripts...
[cont-finish.d] done.
[s6-finish] waiting for services.
[s6-finish] sending all processes the TERM signal.
[s6-finish] sending all processes the KILL signal and exiting.
Tried this on the current release of V2 (roflcoopter/viseron:modularize) and it works as expected for me with this config:
ffmpeg:
camera:
camera_1:
name: Camera 1
......
recorder:
codec: h264
filter_args:
- -vf
- transpose=2
No crashes or anything. What do you think, can i close this one?
Trying out the v2 and I got my two cameras running. But trying to rotate the camera didn't work just yet.
ffmpeg: # <-- component
camera: # <-- domain
camera_one: #
...
name: Dome
audio_codec: aac
height: 2592
width: 1944
fps: 5
filter_args:
- '-vf'
- "transpose=2"
Trying out the v2 and I got my two cameras running. But trying to rotate the camera didn't work just yet.
ffmpeg: # <-- component camera: # <-- domain camera_one: # ... name: Dome audio_codec: aac height: 2592 width: 1944 fps: 5 filter_args: - '-vf' - "transpose=2"
Woa that looks odd! Where is that picture from?
When I add the mog2 or darknet, the video filter I'm using (-vf transpose=2) gets overwritten.
[2022-03-05 13:50:24] [DEBUG ] [viseron.components.ffmpeg.stream.camera_one] - FFmpeg decoder command: ffmpeg_camera_one -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 -stimeout 5000000 -c:v h264_cuvid -rtsp_transport tcp -i rtsp://*****:*****@192.168.0.x:554/h264?username=xxxx&password=xxxx -f segment -segment_time 5 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /segments/camera_one/%Y%m%d%H%M%S.mp4 -vf transpose=2 -filter:v fps=1.0 -f rawvideo -pix_fmt nv12 pipe:1
Config
mog2:
motion_detector:
cameras:
camera_one:
fps: 1