CatmullRom Splines-Based Regression for Image Forgery Localization
Authors: Li Zhang, Mingliang Xu, Dong Li, Jianming Du, Rujing Wang
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments show the superiority of CSR-Net to existing state-of-the-art methods, not only on standard natural image datasets but also on social media datasets. |
| Researcher Affiliation | Academia | 1 Hefei Institute of Physical Science, Chinese Academy of Sciences, China 2 University of Science and Technology of China, China |
| Pseudocode | No | The paper describes the method and its components in text and diagrams, but does not provide structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper discusses the unavailability of source code for a compared method but does not provide concrete access to its own source code for the methodology described. |
| Open Datasets | Yes | Pre-training Data We create a sizable image tampering dataset and use it to pre-train our model. This dataset includes three categories: 1) splicing, 2) copy-move, and 3) removal. Testing Datasets Following (Wang et al. 2022; Hu et al. 2020), we evaluate our model on CASIA (Dong, Wang, and Tan 2013), Columbia (Hsu and Chang 2006), NIST16 (Guan et al. 2019), COVER (Wen et al. 2016). |
| Dataset Splits | No | The paper mentions training on 'training split' and evaluating on 'test split' of datasets, but does not provide specific percentages, sample counts, or explicit details for validation dataset splits needed for reproduction. |
| Hardware Specification | Yes | Implemented by Py Torch, our model is trained with Ge Force GTX 3090, using Adam as the optimizer. |
| Software Dependencies | No | The paper mentions 'Py Torch' as the implementation framework but does not provide specific version numbers for it or any other software dependencies. |
| Experiment Setup | Yes | The input images are resized to 512 × 512. In this work, the backbone network is Res Net-50, pre-trained on Image Net. Implemented by Py Torch, our model is trained with Ge Force GTX 3090, using Adam as the optimizer. ...Ablation experiments (In the ablation analysis part) show that Catmull Rom splines can be reliable for this task when τ is set to 16. ...In the actual experiment, we take k = 3. |