bytetrack-counter-cpp/RUN.md

ByteTrack Counter C++ — How to Run

Build

cd /opt/bytetrack-counter-cpp
mkdir -p build && cd build
cmake .. -DHAS_RKNN=ON
make -j$(nproc)

Run

Run from the project root (looks for .env in CWD by default):

cd /opt/bytetrack-counter-cpp
./build/bytetrack-counter

Specify a custom env file

Pass the path as the first argument:

./build/bytetrack-counter /path/to/custom.env

Or via environment variable:

ENV_FILE=/path/to/custom.env ./build/bytetrack-counter

Priority

1. CLI argument (if provided)
2. ENV_FILE environment variable (if set)
3. ./.env in the working directory

Service (systemd)

sudo systemctl restart bytetrack-counter
journalctl -u bytetrack-counter -f

Environment Variables

See .env.example for a full reference. Key variables:

Variable	Default	Description
SOURCE	rtsp://user:pass@192.168.0.100:554/stream1	RTSP camera URL
MODEL_PATH	/opt/models/yolo.rknn	RKNN model file
IMGSZ	320	Model input size
CONF	0.3	Detection confidence threshold
CORE_MASK	1	RKNN NPU core mask (1, 2, 3, 7)
CAMERA_NAME	CC1	Camera ID label
CROSS_DIRECTION	rtl	Line crossing direction: rtl, ltr, both
LINE_X_FRAC	0.5	Counting line position (fraction of frame width)
TRACK_HIGH_THRESH_0	0.5	ByteTrack high threshold (ayam)
TRACK_MATCH_THRESH_0	0.8	ByteTrack match threshold (ayam)
TRACK_HIGH_THRESH_1	0.5	ByteTrack high threshold (talenan)
TRACK_MATCH_THRESH_1	0.6	ByteTrack match threshold (talenan)
AYAM_CLASS_ID	0	YOLO class ID for ayam
TALENAN_CLASS_ID	1	YOLO class ID for talenan
RATE_WINDOW_SEC	60	Rate calculation sliding window (seconds)
RECONNECT_DELAY_SEC	3	Reconnect delay on stream drop / initial connect
MAX_RECONNECT_ATTEMPTS	0	Max reconnect attempts (0 = unlimited)
LIVE_STREAM_ENABLED	false	Write JPEG snapshot for web dashboard
RECORD_VIDEO	false	Record MP4 video segments
OUTPUT_DIR	/opt/bytetrack-counter-cpp	Data output directory
DB_PATH	{OUTPUT_DIR}/bytetrack_counter.db	SQLite database path