Extend Camera Object Detection for multiple cameras within the same node -> Batched inference, tensorRT

[Kishore-Yogaraj]

Need to use TensorRT, something like https://github.com/noahmr/yolov5-tensorrt for yolov8 ➝ https://github.com/triple-Mu/YOLOv8-TensorRT/blob/main/infer-det.py

will be in C++

also look at https://github.com/NVIDIA-AI-IOT/YOLOv5-with-Isaac-ROS

[justin] read 2.1. Batching section here https://docs.nvidia.com/deeplearning/tensorrt/archives/tensorrt-700/pdf/TensorRT-Best-Practices.pdf Approach: There are three subtasks for this:

• Create a node to synchronize multiple camera messages to publish a batched_camera_message
• Research how to convert yolov8 to tensorrt and modify the model output to support batched inference (see https://medium.com/@DeeperAndCheaper/yolov8-batch-inference-implementation-using-tensorrt-2-converting-to-batch-model-engine-e02dc203fc8b )
• Create a node that subscribes to the batched_camera_message and runs inference on the batch of images

[Arfan12630]

Buffer for getting messages from all 6 camera angles https://wiki.ros.org/message_filters#registerCallback.28.29