This adds two packages, object_detection and scene_graph. The object detection package is configured with a python package, and a test via pytest. We also configure a single dummy message that is compiled at build time. The tests just verify basic cv_bridge and msg publishing.
$ docker compose run app catkin test
Starting >>> object_detection
Starting >>> scene_graph
Output << scene_graph:results /app/logs/scene_graph/test.results.000.log
Summary: 0 tests, 0 errors, 0 failures, 0 skipped
cd /app/build/scene_graph; catkin test --get-env scene_graph | catkin env -si catkin_test_results; cd -
Finished <<< scene_graph [ 0.3 seconds ]
...
[Testcase: testtest_detect] ... ok
[ROSTEST]-----------------------------------------------------------------------
[object_detection.rosunit-test_detect/test_recieve_image_mock][passed]
[object_detection.rosunit-test_detect/test_msg_detected_object_list][passed]
[object_detection.rosunit-test_detect/test_msg_detected_object_list_publish][passed]
SUMMARY
* RESULT: SUCCESS
* TESTS: 3
* ERRORS: 0
* FAILURES: 0
rostest log file is in /root/.ros/log/rostest-autobots-vm-127.log
cd /app/build/object_detection; catkin test --get-env object_detection | catkin env -si /usr/bin/make run_tests; cd -
Output << object_detection:results /app/logs/object_detection/test.results.000.log
Summary: 4 tests, 0 errors, 0 failures, 0 skipped
cd /app/build/object_detection; catkin test --get-env object_detection | catkin env -si catkin_test_results; cd -
Finished <<< object_detection [ 3.2 seconds ]
[test] Summary: All 2 packages succeeded!
[test] Ignored: None.
[test] Warnings: None.
[test] Abandoned: None.
[test] Failed: None.
[test] Runtime: 3.2 seconds total.
There are a few notable oddities to consider here:
_pycache__ files can cause issues when mounting the local directory into the container. These can be resolved by removing all the compiled bytecode.
Getting a list of all installable packages can be tricky, I've left notes in the docker config.
I think the next thing we should add is yolov8, and to wrap that functionality into a sane "perceiver" (which may be a better name for the object_detection package). We should use a requirements.txt at the package root to specify the dependencies, because ROS does not support the wider pip ecosystem. Thus, we need to rely on external methods to get interesting python packages into the package.
This adds two packages,
object_detectionandscene_graph. The object detection package is configured with a python package, and a test via pytest. We also configure a single dummy message that is compiled at build time. The tests just verify basiccv_bridgeandmsgpublishing.There are a few notable oddities to consider here:
_pycache__files can cause issues when mounting the local directory into the container. These can be resolved by removing all the compiled bytecode.I think the next thing we should add is yolov8, and to wrap that functionality into a sane "perceiver" (which may be a better name for the object_detection package). We should use a
requirements.txtat the package root to specify the dependencies, because ROS does not support the wider pip ecosystem. Thus, we need to rely on external methods to get interesting python packages into the package.