Flow-Based Feature Fusion for Vehicle-Infrastructure Cooperative 3D Object Detection
Authors: Haibao Yu, Yingjuan Tang, Enze Xie, Jilei Mao, Ping Luo, Zaiqing Nie
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results demonstrate that our proposed method outperforms existing cooperative detection methods while only requiring about 1/100 of the transmission cost of raw data and covers all latency in one model on the DAIR-V2X dataset. The code is available at https://github.com/haibao-yu/FFNet-VIC3D. |
| Researcher Affiliation | Collaboration | Haibao Yu1,2, Yingjuan Tang2,3, Enze Xie1, Jilei Mao2, Ping Luo1,4, Zaiqing Nie2 1The University of Hong Kong 2Institute for AI Industry Research (AIR), Tsinghua University 3Beijing Institute of Technology 4Shanghai AI Laboratory |
| Pseudocode | No | The paper describes the model architecture and processes in text and diagrams, but it does not include any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code is available at https://github.com/haibao-yu/FFNet-VIC3D. |
| Open Datasets | Yes | We used public and real-world DAIR-V2X dataset (40), which comprises over 100 scenes and 18,000 data pairs captured from infrastructure and vehicle sensors (Cameras and Li DARs) at 28 challenging traffic intersections. |
| Dataset Splits | Yes | The dataset is divided into train/val/test sets in a 5:2:3 ratio, with all models evaluated on the val set. |
| Hardware Specification | Yes | All training and evaluation were performed on an NVIDIA Ge Force RTX 3090 GPU. |
| Software Dependencies | No | We utilized MMDetection3D (1) as our codebase. While 'MMDetection3D' is named, no specific version number for this or any other software dependency (like Python, PyTorch, CUDA versions) is provided in the text. |
| Experiment Setup | Yes | We trained the feature fusion base model on the DAIR-V2X training set for 40 epochs, with a learning rate of 0.001 and weight decay of 0.01. ... We trained the feature flow generator on Du for 10 epochs with a learning rate of 0.001 and weight decay of 0.01. |