Real-time Stereo-based 3D Object Detection for Streaming Perception
Authors: Changcai Li, Zonghua Gu, Gang Chen, Libo Huang, Wei Zhang, Huihui Zhou
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on the KITTI Tracking dataset show that, compared with the strong baseline, Stream DSGN significantly improves the streaming average precision by up to 4.33%. Our code is available at https://github.com/weiyangdaren/stream DSGN-pytorch. |
| Researcher Affiliation | Academia | Changcai Li1,2 Zonghua Gu3 Gang Chen1,2, Libo Huang4 Wei Zhang2 Huihui Zhou2 1Sun Yat-sen University 2Pengcheng Laboratory 3Hofstra University 4National University of Defense Technology |
| Pseudocode | No | The paper does not contain any sections or figures explicitly labeled 'Pseudocode' or 'Algorithm'. |
| Open Source Code | Yes | Our code is available at https://github.com/weiyangdaren/stream DSGN-pytorch. ... Our code is temporarily hosted on an anonymous platform: https:// anonymous.4open.science/r/stream DSGN-FD29.. If the paper is accepted, we will release the source code on Git Hub. |
| Open Datasets | Yes | We conduct our experiments on KITTI tracking dataset [10], which provides stereo images and a higher frame rate (10Hz). ... Andreas Geiger, Philip Lenz, and Raquel Urtasun. Are we ready for autonomous driving? the kitti vision benchmark suite. In 2012 IEEE conference on computer vision and pattern recognition, pages 3354 3361. IEEE, 2012. |
| Dataset Splits | Yes | We partition the training set with 20 scenes into numerous frame sequences, each comprising 40 frames. Among them, the even-numbered sequences are used for training with a total of 4,291 frames, and the odd-numbered sequences are used for testing with a total of 3,672 frames. ... our split tracking dataset, which consists of 4,291 training samples and 3,672 validation samples |
| Hardware Specification | Yes | Our experiments are conducted on the NVIDIA TITAN RTX platform with a total of 20 epochs. |
| Software Dependencies | No | The paper mentions 'Adam optimizer [23]' and 'One Cycle learning rate decay strategy [47]', but does not provide specific version numbers for these or other software dependencies like Python, PyTorch, or CUDA. |
| Experiment Setup | Yes | Our experiments are conducted on the NVIDIA TITAN RTX platform with a total of 20 epochs. During the training phase, the Adam optimizer [23] is employed in conjunction with the One Cycle learning rate decay strategy [47]. The initial learning rate is set to 1e 3, progressively increased to 1e 2, and finally decayed to 1e 8. ... τ = 0.8. ... λ = 0.5. |