Configuration Options

Visit the Configuration File section for information about how to use these configuration options.

Workspace

This is the location where DTrack stores data. This includes saved recordings, audio clips, trained models, temporary data, etc.

Default Value: "_workspace/"

Type	Configuration Variable	Environment Variable
string	workspace	DTRACK_WORKSPACE

Workspace Keep Temp

Keep temporary data after extracted for review.

Default Value: false

Type	Configuration Variable	Environment Variable
boolean	keep_temp	DTRACK_KEEP_TEMP

Record Audio Device

The device identifier used to record audio.

Default Value: "plughw"

Type	Configuration Variable	Environment Variable
string	audio_device	RECORD_AUDIO_DEVICE

• List audio devices: ffmpeg -loglevel warning -sources alsa

  • Sample Output:

    Auto-detected sources for alsa:
    null [Discard all samples (playback) or generate zero samples (capture)]
    hw:CARD=Snowflake,DEV=0 [Direct hardware device without any conversions]
    plughw:CARD=Snowflake,DEV=0 [Hardware device with all software conversions]
    default:CARD=Snowflake [Default Audio Device]
    sysdefault:CARD=Snowflake [Default Audio Device]
    front:CARD=Snowflake,DEV=0 [Front output / input]
    dsnoop:CARD=Snowflake,DEV=0 [Direct sample snooping device]
    

  • Look for devices with: "software conversions"

  • It is likely worth taking time to test available options.

Record Audio Options

Additional ffmpeg options used for audio capture.

Default Value: [ "-f", "alsa" ]

Type	Configuration Variable	Environment Variable
list	audio_options	RECORD_AUDIO_OPTIONS

Record Video Device

The device identifier used to record video.

Default Value: "/dev/video0"

Type	Configuration Variable	Environment Variable
string	video_device	RECORD_VIDEO_DEVICE

• List video devices: ls -1 /dev/video*

  • Sample Output:

    /dev/video0
    /dev/video1
    /dev/video2
    /dev/video3
    

• List capabilities: ffmpeg -hide_banner -f v4l2 -list_formats all -i <DEVICE>

  • Sample Output:

    [v4l2 @ 0x5612] Compressed:       mjpeg: Motion-JPEG : 640x360 640x480 960x540 1024x576 1280x720 1920x1080 2560x1440 3840x2160
    [v4l2 @ 0x5612] Compressed:        h264:       H.264 : 640x360 640x480 960x540 1024x576 1280x720 1920x1080 2560x1440 3840x2160
    [v4l2 @ 0x5612] Compressed: Unsupported:        HEVC : 640x360 640x480 960x540 1024x576 1280x720 1920x1080 2560x1440 3840x2160
    [v4l2 @ 0x5612] Raw       :     yuyv422:  YUYV 4:2:2 : 640x360 640x480 960x540 1024x576 1280x720 1920x1080
    [in#0 @ 0x5612] Error opening input: Immediate exit requested
    Error opening input file /dev/video2.
    

  • This output shows a camera capable of recording 4K video.

Record Video Options

Additional ffmpeg options used for video capture.

List video devices using: ls /dev/video*

List video device capabilities using: v4l2-ctl -d $DEVICE --list-formats-ext; look for 'H264' (H.264, compressed).

Default Value: [ "-f", "v4l2", "-input_format", "h264", "-video_size", "1920x1080", "-framerate", "20" ]

Type	Configuration Variable	Environment Variable
list	video_options	RECORD_VIDEO_OPTIONS

Record Video Timestamp

Long string that defines the datestamp that is added on top of recorded video.

The default value uses Monospace Bold, set to a bright Red, 48-Point font. This font is provided by the fonts-freefont-ttf package,

This filter is applied to every frame, so it is possible to add hflip,vflip, to the start of the string, which will result in an image that is flipped both horizontally and vertically--perfect for an upside-down camera.

Default Value: "drawtext=fontfile=/usr/share/fonts/truetype/freefont/FreeMonoBold.ttf:text=%{localtime}:fontcolor=red@0.9:x=7:y=7:fontsize=48"

Type	Configuration Variable	Environment Variable
string	video_timestamp	RECORD_VIDEO_TIMESTAMP

Record Video Advanced

Video creation arguments that ultimately control the final recorded video quality.

A sensible range for 1080p video is between 3M and 8M, with 3.5M being the observed upper limit on a Raspberry Pi 5--due to lack of hardware encoder.

Default Value: [ "libx264", "-crf", "23", "-preset", "fast", "-tune", "zerolatency", "-maxrate", "3M", "-bufsize", "24M" ]

Type	Configuration Variable	Environment Variable
list	video_advanced	RECORD_VIDEO_ADVANCED

Record Inspect Models

List of tags used for training, and trained models used for automatic detection.

Default Value: [ ]

Type	Configuration Variable	Environment Variable
list	inspect_models	RECORD_INSPECT_MODELS

Record Inspect Backlog

Defines the maximum number of "segments" that can back up in each queue before newly recorded audio is discarded from queue.

Default Value: 5

Type	Configuration Variable	Environment Variable
integer	inspect_backlog	RECORD_INSPECT_BACKLOG

Record Inspect Trust

Confidence level required to assume a match was found.

Default Value: 0.50

Type	Configuration Variable	Environment Variable
decimal	inspect_trust	RECORD_INSPECT_TRUST

Record Duration

Length of time to use for each recording.

• Very short recordings will suffer inadequate compression and are harder to review.
• Very long recordings will be much larger and more difficult to submit.

Default Value: "00:10:00"

Type	Configuration Variable	Environment Variable
string	record_duration	RECORD_DURATION

Train Batch Size

Number of samples to use in each round of training.

Default Value: 16

Type	Configuration Variable	Environment Variable
integer	train_batch_size	TRAIN_BATCH_SIZE

Train Epochs

Maximum number of learning attempts before stopping.

Default Value: 200

Type	Configuration Variable	Environment Variable
integer	train_epochs	TRAIN_EPOCHS

Train Patience

Maximum number of attempts to build a better model before stopping early.

Default Value: 10

Type	Configuration Variable	Environment Variable
integer	train_patience	TRAIN_PATIENCE

Train Rate

Step size used to adjust learning. This may be decreased during training to prevent over-fitting data samples.

Default Value: 0.0001

Type	Configuration Variable	Environment Variable
decimal	train_rate	TRAIN_RATE