final presentation contents

• About V2X-Seq-TFD and basline and result (HiVT, v2itrajnet, v2x-graph)
  • V2X-Seq-TFD: https://arxiv.org/pdf/2305.05938
  • HiVT: https://openaccess.thecvf.com/content/CVPR2022/papers/Zhou_HiVT_Hierarchical_Vector_Transformer_for_Multi-Agent_Motion_Prediction_CVPR_2022_paper.pdf
  • v2x-graph: https://arxiv.org/pdf/2311.00371
  • v2itrajnet: https://koreascience.kr/article/CFKO202532543238938.pub?&lang=ko
• how to auto-calibration in monocular camera (cheonan cctv)?
  • auto-calibration: https://openaccess.thecvf.com/content/ICCV2021/papers/Revaud_Robust_Automatic_Monocular_Vehicle_Speed_Estimation_for_Traffic_Surveillance_ICCV_2021_paper.pdf
• dataset construction pipeline
  • include detection, calculate velocity, target_agent, hd-map
  • detection: https://huggingface.co/mshamrai/yolov8x-visdrone
• baseline performance (hivt, modified v2x, v2itrajnet) and ablation (w/o map)
  • how i modify v2x graph
  • quantitative, qualitative result

feedback

• [x] v2x-graph 변경 사항 그림 그리기
• [ ] velocity 전파 알고리즘 추가하기
• [x] interestingness score figure 직관적으로 그려보기