I think TensorRT is active. I also for testing purposes decreased the image inference size but have to pay attention that it does not get to small because I have wealded seams to detect on the hoses and for that only few pixels. Therefore the challenge keeping as big as possible…
The reduction of inference size did not help it is the same and also MAXN is active
ros2 topic hz /hose_grasp/image
[hose_grasp_node-4] [INFO] [1779997179.855305411] [hose_grasp_node]: Standard-Modell geladen (kein Jetson-Profil)
[hose_grasp_node-4] [INFO] [1779997180.240573131] [hose_grasp_node]: Lade TRT-Backbone aus Cache…
[hose_grasp_node-4] /workspaces/isaac_ros-dev/build/hose_grasp_detection/hose_grasp_detection/hose_grasp_node.py:247: FutureWarning: You are using torch.load with weights_only=False (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See pytorch/SECURITY.md at main · pytorch/pytorch · GitHub for more details). In a future release, the default value for weights_only will be flipped to True. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via torch.serialization.add_safe_globals. We recommend you start setting weights_only=True for any use case where you don’t have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
[hose_grasp_node-4] backbone_trt.load_state_dict(torch.load(trt_cache))
[hose_grasp_node-4] [05/28/2026-21:39:40] [TRT] [W] Using an engine plan file across different models of devices is not recommended and is likely to affect performance or even cause errors.
[hose_grasp_node-4] [INFO] [1779997180.636613039] [hose_grasp_node]: TRT-Backbone geladen.
[hose_grasp_node-4] [INFO] [1779997180.638427619] [hose_grasp_node]: Modell | Device: cuda:0 | Klassen (8): [‘hose_1_inch’, ‘hose_end’, ‘weld_seam’, ‘gripper_bake’, ‘hose_holder’, ‘tube_holder_biowelder’, ‘blade’] | Hose: {1} | Forbidden: {2, 3} | Inference: 512px
[hose_grasp_node-4] [INFO] [1779997180.679989873] [hose_grasp_node]: HoseGraspNode bereit.
[hose_grasp_node-4] [05/28/2026-21:39:40] [TRT] [W] Using default stream in enqueueV3() may lead to performance issues due to additional calls to cudaStreamSynchronize() by TensorRT to ensure correct synchronization. Please use non-default stream instead.
[hose_grasp_node-4] [INFO] [1779997196.994144924] [hose_grasp_node]: Intrinsics: fx=265.0 fy=265.0 cx=316.1 cy=173.9