As a dataflow author,
I want dora-zed to publish only camera data,
so that the node has no inference dependency and builds on any platform the ZED SDK supports.
Requirements
Delete src/engines/ in full: engine.h, zed_native.cc, zed_native.h, yolo_onnx.cc, yolo_onnx.h, yolo_helpers.h.
Remove the objects output and serialize_objects from src/main.cc.
Remove FIELDS_PER_OBJECT and the detector construction and dispatch.
Remove ZED_ENGINE, ZED_DETECTION_MODEL, ZED_DETECTION_CONFIDENCE, ZED_DETECTION_TRACKING and all ZED_ONNX_* variables.
Remove ZED_PC_MODE=ROI; keep FULL and HALF.
Remove the test_yolo_onnx and test_yoloworld_onnx targets from CMakeLists.txt.
Remove find_package(onnxruntime REQUIRED), ONNXRUNTIME_INCLUDE_DIRS, ONNXRUNTIME_LIBRARIES and the onnxruntime linkage.
Delete test/assets/.
Update README.md: remove the detection engine tables, the objects output and its layout, and the ONNX Runtime dependency row.
Acceptance Criteria
The node configures and builds on a machine with no ONNX Runtime installed.
Outputs are image_bgr, point_cloud and camera_info only.
ZED_PC_MODE=ROI crops the point cloud to detector->objects() bounding boxes in src/main.cc and cannot survive the removal.
zed_native is dropped rather than moved to liria/dora-yolo-onnx. It is the ZED SDK's
built-in TensorRT detector and is bound to the live camera handle
(detector->enable(camera.handle()), detector->detect(camera.handle())), so it cannot
run in a node that receives images over a dataflow. The code remains reachable at tag v0.1.0.
Removing detection also resolves the Jetson build constraint: CMakeLists.txt currently
requires ONNX Runtime unconditionally, while README.md states the native engine works
without it and that no official ONNX Runtime GPU binary exists for ARM64.
# Feature: Remove detection from dora-zed
## User Story
As a dataflow author,
I want dora-zed to publish only camera data,
so that the node has no inference dependency and builds on any platform the ZED SDK supports.
## Requirements
- Delete `src/engines/` in full: `engine.h`, `zed_native.cc`, `zed_native.h`, `yolo_onnx.cc`, `yolo_onnx.h`, `yolo_helpers.h`.
- Remove the `objects` output and `serialize_objects` from `src/main.cc`.
- Remove `FIELDS_PER_OBJECT` and the detector construction and dispatch.
- Remove `ZED_ENGINE`, `ZED_DETECTION_MODEL`, `ZED_DETECTION_CONFIDENCE`, `ZED_DETECTION_TRACKING` and all `ZED_ONNX_*` variables.
- Remove `ZED_PC_MODE=ROI`; keep `FULL` and `HALF`.
- Remove the `test_yolo_onnx` and `test_yoloworld_onnx` targets from `CMakeLists.txt`.
- Remove `find_package(onnxruntime REQUIRED)`, `ONNXRUNTIME_INCLUDE_DIRS`, `ONNXRUNTIME_LIBRARIES` and the onnxruntime linkage.
- Delete `test/assets/`.
- Update `README.md`: remove the detection engine tables, the `objects` output and its layout, and the ONNX Runtime dependency row.
## Acceptance Criteria
- [ ] The node configures and builds on a machine with no ONNX Runtime installed.
- [ ] Outputs are `image_bgr`, `point_cloud` and `camera_info` only.
- [ ] `git grep -i -e onnx -e detector -e detection` returns no matches.
- [ ] `README.md` documents no detection capability.
## Notes
`ZED_PC_MODE=ROI` crops the point cloud to `detector->objects()` bounding boxes in
`src/main.cc` and cannot survive the removal.
`zed_native` is dropped rather than moved to `liria/dora-yolo-onnx`. It is the ZED SDK's
built-in TensorRT detector and is bound to the live camera handle
(`detector->enable(camera.handle())`, `detector->detect(camera.handle())`), so it cannot
run in a node that receives images over a dataflow. The code remains reachable at tag
`v0.1.0`.
Removing detection also resolves the Jetson build constraint: `CMakeLists.txt` currently
requires ONNX Runtime unconditionally, while `README.md` states the `native` engine works
without it and that no official ONNX Runtime GPU binary exists for ARM64.
cristhian
added this to the v0.2.0 Pure camera source on dora 1.0 milestone 2026-09-04 20:34:50 +00:00
cristhian
added the chore label 2026-09-04 20:34:50 +00:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Feature: Remove detection from dora-zed
User Story
As a dataflow author,
I want dora-zed to publish only camera data,
so that the node has no inference dependency and builds on any platform the ZED SDK supports.
Requirements
src/engines/in full:engine.h,zed_native.cc,zed_native.h,yolo_onnx.cc,yolo_onnx.h,yolo_helpers.h.objectsoutput andserialize_objectsfromsrc/main.cc.FIELDS_PER_OBJECTand the detector construction and dispatch.ZED_ENGINE,ZED_DETECTION_MODEL,ZED_DETECTION_CONFIDENCE,ZED_DETECTION_TRACKINGand allZED_ONNX_*variables.ZED_PC_MODE=ROI; keepFULLandHALF.test_yolo_onnxandtest_yoloworld_onnxtargets fromCMakeLists.txt.find_package(onnxruntime REQUIRED),ONNXRUNTIME_INCLUDE_DIRS,ONNXRUNTIME_LIBRARIESand the onnxruntime linkage.test/assets/.README.md: remove the detection engine tables, theobjectsoutput and its layout, and the ONNX Runtime dependency row.Acceptance Criteria
image_bgr,point_cloudandcamera_infoonly.git grep -i -e onnx -e detector -e detectionreturns no matches.README.mddocuments no detection capability.Notes
ZED_PC_MODE=ROIcrops the point cloud todetector->objects()bounding boxes insrc/main.ccand cannot survive the removal.zed_nativeis dropped rather than moved toliria/dora-yolo-onnx. It is the ZED SDK'sbuilt-in TensorRT detector and is bound to the live camera handle
(
detector->enable(camera.handle()),detector->detect(camera.handle())), so it cannotrun in a node that receives images over a dataflow. The code remains reachable at tag
v0.1.0.Removing detection also resolves the Jetson build constraint:
CMakeLists.txtcurrentlyrequires ONNX Runtime unconditionally, while
README.mdstates thenativeengine workswithout it and that no official ONNX Runtime GPU binary exists for ARM64.