General repository for TMK Projects Group. This is INTEGRATION repository.
main - history of Releases and Tags
dev - daily build repository
TMKDCL-xxx - developer branches by issues in Jira for merge by pull-request strategy to dev-branch
Prepare your docker image //TODO: or use it from the project repository
//TODO: write how-to
Prepare and run data sources and server (process pipeline adn tcp command server) as docker services
//TODO:
Prepare pipline configuration file
Run the system manager (tcp client //TODO: or send http requessts). You can use netcat utility for this goals.
Start the system using commands by command API //TODO: command specification
You can attach to process for debug //TODO: write how-to //TODO: attach isn't testing now
Obtain the results (logs) of process using docker log ???
//TODO: find the better solution then docker log
!ATTENTION! You CAN build and run this solution in SOS cases directly. Enviroment is prepared for this issue. But it's strongly NOT RECOMMENDED.
Enviroment:
sudo apt install //TODO: lib
Build:
git clone --recurce-submodule https://github.com/tmk-org/TMK_DCL.git
cd TMK_DCL
mkdir -p build && cd build
cmake ..
make -j$(nproc)
Run from build-directory
cd dclcomponents
Start data source
TEST_SOURCE_PATH="/mnt/testdb/pencil_frames_5" bin/tmk-camera 9000
UPD: for usage more then 1 source ports should be 9000 + 10 * i, where i is source number Start server
GLOG_minloglevel=2 bin/tmk-server 4444
PS: (Not tested) you can start tmk-camera and tmk-server as daemon with //TODO: check the work is correct
bin/tmk-server 4444 1
Start command tcp client
bin/tmk-manager localhost 4444
(or use netcat utility for this goals as netcat localhost 4444)
Test system:
- Send
stopcommand to server by tcp. After some waitting the system should be stopped without any errors. - Send
config:defaultcommand for start pipeline with empty middle orconfig:pgcommand for start with connection to the database - You can send
config:demofor string filter pipeline idea demonstration - (Feature: not implemented) Send
initcommand for start empty pipeline, other modules could be attached in pipeline by commands - (Not tested) Send
config:path_to_jsoncommand
//TODO: init command by IP has paramerer - configuration .json-file for configure pipeline -- in progress
//TODO:
Check that all modules are started (by default: tmk-server, tmk-first, tmk-middle, tmk-last)
ps aux | grep tmk
- Send
stopcommand. Pipeline should be stopped correctly. Check that all modules are finished.
Run sender to winnum (database shoould be available)
cd ../winnum
python3 sent_to_winnum.py
Configuration is hard-code now =(
Pipeline configuration: dclprocessor/src/pipeline_demo.c, функция create_default_configuratuion. Важно: количество модулей, их порядок, строка старта. First и last фикстированы. Учесть, что количество модулей = количество объектов + 1 в памяти одновременно.
Capture configuration: dclprocessor/src/collector_description.c, функция read_collector_description. Важно: количество источников, выделяемая память под объекты (в случае, если не хватит - будет падение)
Number frames in object: dclprocessor/include/dclprocessor/source.h, #define OBJECT_DETECTION_FRAME_LIMIT 50 (в минимальном прототипе 100, но хочется укладываться в сумме в 16Гб памяти)
Fake tcp-camera: dclprocessor/include/dclprocessor/cvcamera.h, размер количество кадров в датасете, который "стримим". По умолчанию dataset "pencil_frames_5", первые 10 кадров по циклу