MGQFormer: Mask-Guided Query-Based Transformer for Image Manipulation Localization

Authors: Kunlun Zeng, Ri Cheng, Weimin Tan, Bo Yan

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on multiple benchmarks show that our method significantly improves over state-of-the-art methods. Experiment Experiment Setup Testing Datasets. We first pre-train our model with the dataset synthesized by PSCC-Net (Liu et al. 2022). Then we evaluate our model on CASIA dataset (Dong, Wang, and Tan 2013), Columbia dataset (Hsu and Chang 2006), NIST16 dataset (Guan et al. 2019) and IMD20 dataset (Novozamsky, Mahdian, and Saic 2020).
Researcher Affiliation Academia Kunlun Zeng*, Ri Cheng*, Weimin Tan , Bo Yan School of Computer Science, Shanghai Key Laboratory of Intelligent Information Processing, Fudan University klceng22@m.fudan.edu.cn, rcheng22@m.fudan.edu.cn, wmtan@fudan.edu.cn, byan@fudan.edu.cn
Pseudocode No No pseudocode or clearly labeled algorithm block was found in the paper.
Open Source Code No The paper does not provide any statement or link indicating that the source code for the methodology is openly available.
Open Datasets Yes Experiment Experiment Setup Testing Datasets. We first pre-train our model with the dataset synthesized by PSCC-Net (Liu et al. 2022). Then we evaluate our model on CASIA dataset (Dong, Wang, and Tan 2013), Columbia dataset (Hsu and Chang 2006), NIST16 dataset (Guan et al. 2019) and IMD20 dataset (Novozamsky, Mahdian, and Saic 2020).
Dataset Splits Yes As shown in Figure 3, we display the AUC (%) scores on the validation split of the synthesized dataset during training.
Hardware Specification Yes The MGQFormer is implemented on the Pytorch with an NVIDIA GTX 1080 Ti GPU.
Software Dependencies No The paper mentions 'Pytorch' but does not specify a version number or other software dependencies with specific versions.
Experiment Setup Yes Implementation Details. The MGQFormer is implemented on the Pytorch with an NVIDIA GTX 1080 Ti GPU. All input images are resized to 384 x 384. We use Adam as the optimizer, and the learning rate decays from 2.5e-7 to 1.5e-8 with a batch size of 2.