Enhancing Motion Deblurring in High-Speed Scenes with Spike Streams
Authors: Shiyan Chen, Jiyuan Zhang, Yajing Zheng, Tiejun Huang, Zhaofei Yu
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The experimental results demonstrate that our method effectively recovers clear RGB images from highly blurry scenes and outperforms state-of-the-art deblurring algorithms in multiple settings. Furthermore, we build two extensive synthesized datasets for training and validation purposes, encompassing high-temporal-resolution spikes, blurry images, and corresponding sharp images. |
| Researcher Affiliation | Academia | Shiyan Chen1,2 Jiyuan Zhang1,2 Yajing Zheng1,2 Tiejun Huang1,2,3 Zhaofei Yu1,2,3 1School of Computer Science, Peking University 2National Key Laboratory for Multimedia Information Processing, Peking University 3Institute for Artificial Intelligence, Peking University |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any explicit statement about releasing the source code or a link to a code repository. |
| Open Datasets | Yes | Therefore, we built two large-scale spike-based datasets based on existing real image datasets [39, 46] for training and validation. The first one, named Spk-X4K1000FPS , is built on the dataset X4K1000FPS which contains clear images captured by 1000fps cameras. The second one, named Spk-Go Pro is built on the Go Pro [39] dataset in a similar way as above. |
| Dataset Splits | No | We build two extensive synthesized datasets for training and validation purposes. However, specific percentages, sample counts, or explicit splitting methodologies for these sets are not provided in the main text. |
| Hardware Specification | No | The paper does not specify any hardware details like GPU models, CPU types, or memory used for running the experiments. It only mentions 'More implementation details are attached in the supplementary materials,' but these are not provided in the main text. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers. It mentions 'More implementation details are attached in the supplementary materials,' but these are not provided in the main text. |
| Experiment Setup | Yes | For Spk-X4K1000FPS, we generate blurry images by setting an exposure window of e = 33 as well as an extreme exposure window of e = 65. ... The loss function of the entire network can be formulated as follows: L = It ˆIt 1 + λ1 Ig t ˆIg t 1 + λ2 It It 1, where λ1 and λ2 are hyperparameters that control the loss terms. All comparative methods were trained from scratch on the proposed dataset. |