Zero-Shot Event-Intensity Asymmetric Stereo via Visual Prompting from Image Domain
Authors: Hanyue Lou, Jinxiu (Sherry) Liang, Minggui Teng, Bin Fan, Yong Xu, Boxin Shi
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on real-world datasets demonstrate the superior zero-shot evaluation performance and enhanced generalization ability of our method compared to existing approaches. |
| Researcher Affiliation | Academia | Hanyue Lou#1,2 Jinxiu Liang#1,2 Minggui Teng1,2 Bin Fan3 Yong Xu4 Boxin Shi 1,2 1 State Key Laboratory of Multimedia Information Processing, School of Computer Science, Peking University 2 National Engineering Research Center of Visual Technology, School of Computer Science, Peking University 3 National Key Laboratory of General AI, School of Intelligence Science and Technology, Peking University 4 School of Computer Science and Engineering, South China University of Technology |
| Pseudocode | No | The paper does not contain any pseudocode or a clearly labeled algorithm block. |
| Open Source Code | Yes | Code available: https://github.com/HYLZ-2019/ZEST |
| Open Datasets | Yes | We evaluate the proposed ZEST framework on the widely-used benchmark dataset for event-intensity stereo matching, the DSEC dataset [13]... To assess generalization, we also evaluate on the MVSEC [45] and M3ED [2] datasets. |
| Dataset Splits | No | Specifically, it provides high-resolution (640 480) stereo event streams captured in outdoor driving scenes using Prophesee Gen 3.1 event cameras across 53 outdoor driving scenarios under diverse lighting. Without specification, all 41 sequences (5 Interlaken sequences, 1 Thun sequence, and 35 Zurich City sequences) in the training set are adopted for evaluation, as the official test split lacks ground truth disparity. |
| Hardware Specification | Yes | All models are tested on an Intel i7-13700K CPU and a single NVIDIA RTX 4090 GPU. |
| Software Dependencies | No | The paper mentions 'Py Torch' but does not specify version numbers for it or any other software dependencies. |
| Experiment Setup | Yes | We optimize this function using gradient descent with the Adam optimizer in Py Torch, running 500 iterations per image. ... We adopt CR and DS for the stereo models, and Depth Anything (DA, checkpoint Depth-Anything-Large, 335.3M parameters) [37] and Mi Da S (Mi, checkpoint BEi T-L-512, 345M parameters) [25] for monocular depth estimation. |