JetsonNanoTracking

Runtime model artifacts used by the Jetson Nano person-tracking tests in the MyECOTracker / JetsonNanoTracking project.

Included files

• runtime_test_models/tracker/resnet18_vggmconv1_otb_dual_large_fp16.engine
  • TensorRT FP16 engine used by the real-video tracking profile.
  • Runtime parameter chain: verified_otb936_run_update -> jetson_fast_trt_rgb_run_update -> jetson_fast_trt_dual_acc
• runtime_test_models/detector/yolo_person_only.torchscript
  • TorchScript person detector used by the dualacc_yolo_ts video runs.

What this repo is for

This repo stores the exact runtime artifacts used on Jetson Nano for three real person-tracking test videos:

• jetson/video_inputs/test1/1.mp4
• jetson/video_inputs/test2/2.mp4
• jetson/video_inputs/test3/3.mp4

The tracker used in those runs is logged as verified_otb936_run_update, which resolves to the dual_acc runtime configuration backed by:

• resnet18_vggmconv1_otb_dual_large_fp16.engine

For runs with detector-assisted correction (dualacc_yolo_ts), the detector artifact is:

• yolo_person_only.torchscript

Notes

• This is an artifact repository, not a standalone Hugging Face Transformers model.
• The TensorRT engine is hardware/runtime specific and intended for Jetson deployment.
• To reproduce the original runs, use the matching project code and parameter aliases from JetsonNanoTracking.