ZoomNet: Part-Aware Adaptive Zooming Neural Network for 3D Object Detection
Authors: Zhenbo Xu, Wei Zhang, Xiaoqing Ye, Xiao Tan, Wei Yang, Shilei Wen, Errui Ding, Ajin Meng, Liusheng Huang12557-12564
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on the popular KITTI 3D detection dataset indicate Zoom Net surpasses all previous state-of-the-art methods by large margins (improved by 9.4% on APbv (Io U=0.7) over pseudo-Li DAR). Ablation study also demonstrates that our adaptive zooming strategy brings an improvement of over 10% on AP3d (Io U=0.7). |
| Researcher Affiliation | Collaboration | Zhenbo Xu,1 Wei Zhang,2 Xiaoqing Ye,2 Xiao Tan,2 Wei Yang,*1 Shilei Wen,2 Errui Ding,2 Ajin Meng,1 Liusheng Huang1 1University of Science and Technology of China 2Department of Computer Vision Technology (VIS), Baidu Inc., China |
| Pseudocode | No | The paper describes its methodology in prose and figures, but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | Both the KFG dataset and our codes will be publicly available at https://github.com/detectRecog/ZoomNet. |
| Open Datasets | Yes | We evaluate our method on the challenging KITTI object detection benchmark (Geiger et al. 2013). |
| Dataset Splits | Yes | we split the 7481 training images into training set and validation set with roughly the same amount. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments. |
| Software Dependencies | No | The paper mentions software components like "PSMnet" but does not specify version numbers for any software dependencies, which are required for reproducibility. |
| Experiment Setup | Yes | In stage (B), the input image size (W, H) is set to (256, 128). E is set to 500 in both training and testing phases... a lower weight w = 0.1 is assigned to the loss of part location and mask. The learning rate in pose estimation and refinement is set to 0.1 of the global learning rate. |