Use Your Own Videos#

In standalone (pymavlink) mode each camera loops a video file from resources/videos/. This page covers converting your own footage into the format the pipelines expect and pointing a camera at it. UAV Mission Compute SDK mode reads live RTSP streams from the SDK instead; replace the SDK’s simulated camera source per that project’s documentation.

The bundled files#

File

Default camera (configs/cameras-pymavlink.yaml)

resources/videos/uav_sample_front.avi

front_camera

resources/videos/uav_sample_bottom.avi

bottom_camera

resources/videos/uav_sample_back.avi

back_camera

A camera’s video: field names any file in resources/videos (.avi, .mp4, .mkv or .ts; no sub-folders). The folder is mounted read-only in the container, so a file added on the host is visible without a restart.

Convert a video#

scripts/convert_video.sh turns anything ffmpeg can read into the format the pipelines are tuned for: H.264 (High profile, yuv420p) in an AVI container, 30 fps, no audio, about 4 Mbps, a keyframe every 250 frames, and trimmed to whole 250-frame groups so the loop point lands on a keyframe. It uses the host’s ffmpeg if installed, otherwise ffmpeg from the pipeline server image, so no extra install is needed.

./scripts/convert_video.sh /path/to/my_clip.mp4 -o resources/videos/my_clip.avi

When replacing several cameras’ files, give them the same frame count so their loops stay in step (a multiple of 250, no longer than the shortest clip):

./scripts/convert_video.sh front.mp4  -o resources/videos/uav_sample_front.avi  -n 1750 -f
./scripts/convert_video.sh bottom.mkv -o resources/videos/uav_sample_bottom.avi -n 1750 -f
./scripts/convert_video.sh back.mov   -o resources/videos/uav_sample_back.avi   -n 1750 -f

The script verifies its output and ends with Verified: valid H.264/AVI, starts on a keyframe, decodes without errors.

Option

Default

Purpose

-o, --output <file.avi>

<input name>.avi

Output file

-n, --frames <N>

whole 250-frame groups

Write exactly N frames (same N for every camera)

-b, --bitrate <rate>

4M

Target bitrate (4M, 4000k)

-r, --fps <n>

30

Output frame rate

-s, --size <WxH>

source size

Scale and letterbox, e.g. 1920x1080

-g, --gop <n>

250

Keyframe interval

-p, --preset <name>

medium

x264 preset (ultrafast … veryslow)

--no-trim

off

Keep every frame instead of trimming to whole groups

-f, --force

off

Overwrite an existing output

Apply it#

Either overwrite a bundled file (keeping its name, as in the example above) or set video: for the camera:

cameras:
  - id: front_camera
    video: my_clip.avi
make check-cameras      # reports "video not found" if the name is wrong
make start-rtsp         # or disarm/arm; the file is re-read at every arming

A pipeline that is already running keeps the file it opened; it picks up a replaced file when it is restarted (next arm, or Manage Pipelines — Stop). For a pipeline started by hand, pass the file name in parameters.video.

Verify#

ffplay -rtsp_transport tcp rtsp://<HOST_IP>:8555/uav-front-camera

Troubleshooting#

scripts/convert_video.sh error messages, video not found from make check-cameras, a pipeline that errors on an unconverted file, and a stream that still shows the old footage are covered in Troubleshooting — Cameras & Video. A soft picture means a low-bitrate source; conversion cannot improve it.