Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Far3D: Expanding the Horizon for Surround-View 3D Object Detection
Authors: Xiaohui Jiang, Shuailin Li, Yingfei Liu, Shihao Wang, Fan Jia, Tiancai Wang, Lijin Han, Xiangyu Zhang
AAAI 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Significantly, Far3D demonstrates So TA performance on the challenging Argoverse 2 dataset, covering a wide range of 150 meters, surpassing several Li DAR-based approaches. The code is available at https://github.com/megvii-research/Far3D. |
| Researcher Affiliation | Collaboration | 1Beijing Institute of Technology 2MEGVII Technology |
| Pseudocode | No | The paper does not contain any pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code is available at https://github.com/megvii-research/Far3D. |
| Open Datasets | Yes | We use the large-scale Argoverse 2 dataset (Wilson et al. 2023) and nu Scenes dataset (Caesar et al. 2020) to explore and evaluate the effectiveness of our approach. |
| Dataset Splits | Yes | Argoverse 2 is a dataset for perception and prediction studies in autonomous driving domain. It contains 1000 scenes with 15 seconds duration and 10Hz annotation frequency. And these total scenes are divided into 700 for training, 150 for validation, and 150 for testing. |
| Hardware Specification | No | The paper mentions backbone architectures (Vo VNet99, Vi T-L, Res Net101) but does not specify the actual hardware (e.g., GPU model, CPU type) used for experiments. |
| Software Dependencies | No | The paper mentions various methods and models (YOLOX, FCOS3D, Stream PETR) and optimizers (Adam W) but does not provide specific software versions (e.g., Python 3.x, PyTorch x.x) for reproducibility. |
| Experiment Setup | Yes | We use Adam W (Loshchilov and Hutter 2017) optimizer with a weight decay of 0.01. The total batch size is 8 and the learning rate is set to 2e-4. The models are totally trained for 6 epochs, following the previous method (Chen et al. 2023). |