Privately Owned Vehicle Work Group Meeting - 2025/04/21 - Slot 1

[m-zain-khawaja]

Agenda

• AutoSeg Production-Ready Optimizations
• Update on EgoPath training
• Update on EgoLanes dataset curation
• Vision-Radar Fusion

---

AutoSeg Production-Ready Optimizations

@m-zain-khawaja :

SceneSeg as well as Scene3D have been exported to ONNX format at FP32 Precision. The traced version of these networks is also available alongside the vanilla Pytorch *.pth file. Scripts for conversion and tracing were moved to Models/exports/export_utils and the tracing script was refactored to account for multiple networks, not just SceneSeg. Documentation was updated to reflect the code changes and links to download the SceneSeg and Scene3D ONNX and Traced models were added to the Models folder

---

EgoPath Network training

• Data Loader class for EgoPath (completed by @TranHuuNhatHuy)
• Augmentations Class for EgoPath (completed by @TranHuuNhatHuy)
• Trainer - EgoPath Network (draft completed by @dahhmani - visualization and testing functions need to be added, Trainer class is ready for testing with full pipeline)
• Main Training Script - draft completed by @TranHuuNhatHuy, full pipeline testing underway
• Data audit - EgoPath Dataset - assigned to multiple open source contributors
  • CurveLanes dataset - @TranHuuNhatHuy (Tran) - in progress
  • ROADWork datset - @sarun-hub (Sarun) - completed
  • CULane dataset - @auo355 (Augustine) - completed
  • BDD100K dataset - @aadarshkt (Aadarsh) - in progress
  • Comma2k19 dataset - @pranavreddy23 (Pranav) - completed
• Network Training - EgoPath Network - assigned to @TranHuuNhatHuy , @sarun-hub , @devang-marvania , @docjag , Mahmoud , Augustine - @m-zain-khawaja has created a number of experiments which can be assigned to open-source contributors - On track to start training this wee week (21st April 2025)

---

EgoLanes Dataset Curation Update

• TuSimple - 6,394 training data samples
• CULane - 15,800 training data samples
• CurveLane - 18,300 training data samples
• BDD100K - complted by @aadarshkt - PR to be reviewd by @TranHuuNhatHuy

---

Vision-Radar Fusion

@m-zain-khawaja :

@docjag has been assigned the task to begin work on developing the Vision-Radar Fusion pipeline, beginning with fusing SceneSeg foreground object labels with Radar detections from automotive Radar. Details about this task can be found here

---

Attendees

• @m-zain-khawaja (Senior Tech Lead at AWF and Privately Owned Vehicle Work Group Chair)
• Narendar Selvaraj (Senior Member of Technical Staff at AMD)
• @mitsudome-r (Chair of Autoware Technical Steering Committee)
• @TranHuuNhatHuy (Masters Robotics Student at Tokyo Institute of Technology)
• @sarun-hub (Undergraduate student at Nagoya University)
• @auo355 (Canada Revenue Agency, Information Technology Analyst/Developer)
• Anthony Chen (ITRI Taiwan Researcher)

Zoom Meeting Video Recording

Video Meeting Link
Please contact the work group lead (@m-zain-khawaja) to request access to a recording of this meeting.