TGIF2: Extended Text-Guided Inpainting Forgery Dataset & Benchmark

Hannes Mareen; Dimitrios Karageorgiou; Paschalis Giakoumoglou; Peter Lambert; Symeon Papadopoulos; Glenn Van Wallendael

cs.CVMar 30, 2026

TGIF2: Extended Text-Guided Inpainting Forgery Dataset & Benchmark

Authors: Hannes Mareen, Dimitrios Karageorgiou, Paschalis Giakoumoglou, Peter Lambert, Symeon Papadopoulos, Glenn Van Wallendael

Organizations: IDLab, Ghent University – imec, Gent, Belgium. · Information Technologies Institute, CERTH, Thessaloniki, Greece.

Abstract

Generative AI has made text-guided inpainting a powerful image editing tool, but at the same time a growing challenge for media forensics. Existing benchmarks, including our text-guided inpainting forgery (TGIF) dataset, show that image forgery localization (IFL) methods can localize manipulations in spliced images but struggle not in fully regenerated (FR) images, while synthetic image detection (SID) methods can detect fully regenerated images but cannot perform localization. With new generative inpainting models emerging and the open problem of localization in FR images remaining, updated datasets and benchmarks are needed. We introduce TGIF2, an extended version of TGIF, that captures recent advances in text-guided inpainting and enables a deeper analysis of forensic robustness. TGIF2 augments the original dataset with edits generated by FLUX.1 models, as well as with random non-semantic masks. Using the TGIF2 dataset, we conduct a forensic evaluation spanning IFL and SID, including fine-tuning IFL methods on FR images and generative super-resolution attacks. Our experiments show that both IFL and SID methods degrade on FLUX.1 manipulations, highlighting limited generalization. Additionally, while fine-tuning improves localization on FR images, evaluation with random non-semantic masks reveals object bias. Furthermore, generative super-resolution significantly weakens forensic traces, demonstrating that common image enhancement operations can undermine current forensic pipelines. In summary, TGIF2 provides an updated dataset and benchmark, which enables new insights into the challenges posed by modern inpainting and AI-based image enhancements. TGIF2 is available at https://github.com/IDLabMedia/tgif-dataset.

Figures & tables

Panel of Figure 1 — Figure 1: Examples of the two ways of inpainting. An (a) authentic image can be inpainted in a (b) selected region as mask, with skis as prompt. In the GenAI-based inpainting process (here using SDXL), (c) the full image is regenerated during editing. To minimize artifacts, (d) only the region corresponding to the mask can be spliced into the authentic image. (e) and (f) provide zoomed-in versions to more clearly show the differences between the original or spliced pixels, on the one hand, and the regenerated pixels outside of the masked area, on the other. Subtle differences due to the regenerative process can be observed in the teeth, sunglasses, beard, etc.

Name	SP	FR	# fake images	Gen. models & pipelines	Beyond obj. repl.
Coco-Glide Guillaro et al. (2023)	✓	-	512	GLIDE	-
AutoSplice Jia et al. (2023)	✓	-	3k	DALL-E2	-
TGIF Mareen et al. (2024)	✓	✓	75k	SD2, SDXL, Adobe Firefly	-
SAGI Giakoumoglou et al. (2025)	✓	✓	95k	SD2, LaMa, HD-Painter, BrushNet, PowerPaint, ControlNet, Inpaint-Anything	-
COinCO Yang et al. (2025)	✓	-	97k	SD2	-
GRE Sun et al. (2024)	✓	-	228k	LaMa, MAT, SD2, ControlNet, PaintByExample, Photoshop	-

Table 1: Overview of datasets with generative inpainted images. SP = spliced, FR = fully regenerated.

Panel of Figure 2 — Figure 2: Side-by-side comparison of (a) a real image of a cat with (b) the mask used for inpainting, and (c)-(h) 6 inpainted versions using 6 different inpainting models used in TGIF2, respectively. Note that other masks can be used as well (see Fig. 3 ).

Panel of Figure 3 — Figure 3: The (a)-(c) three types of masks used and (d)-(f) three corresponding inpainted examples, respectively. All inpainted images used Fig. 2a as input image.

Inpainting Model	Semantic		Random
Inpainting Model	SP	FR	SP	FR
SD2	✓*	✓*	✓	✓
SDXL	-	✓*	-	✓
Adobe Firefly (PS)	✓*	-	-	-
FLUX.1 [schnell]	✓	✓	✓	✓
FLUX.1 [dev]	✓	✓	✓	✓

Table 2: Overview of TGIF2 dataset subsets grouped by inpainting model. Columns indicate whether spliced (SP), fully regenerated (FR), and Semantic or Random mask variants are included.

IFL Method	SD2		PS	FLUX.1						Average
	SD2		PS	[schnell]		[dev]		Fill [dev]		Average
	Sem	Rand	Sem	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	All
PSCC-Net Liu et al. (2022)	0.15	0.17	0.39	0.03	0.09	0.02	0.05	0.08	0.21	0.13	0.13	0.13
SPAN Hu et al. (2020)	0.00	0.00	0.00	0.00	0.00	0.0	0.00	0.00	0.01	0.00	0.00	0.00
ImageForensicsOSN Wu et al. (2022)	0.23	0.09	0.36	0.31	0.11	0.23	0.08	0.12	0.10	0.25	0.09	0.18
MVSS-Net++ Dong et al. (2022)	0.07	0.03	0.08	0.15	0.06	0.11	0.03	0.09	0.09	0.10	0.05	0.08

Table 3: Evaluation of image forgery localization methods on the spliced (SP) subsets of our TGIF2 dataset, for both the semantic (Sem) and random (Rand) versions (F1 score). F1 scores above 0.7 are highlighted in bold.

IFL Method	SD2		SDXL		FLUX.1						Average
	SD2		SDXL		[schnell]		[dev]		Fill [dev]		Average
	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	All
PSCC-Net Liu et al. (2022)	0.05	0.04	0.05	0.09	0.02	0.02	0.02	0.02	0.03	0.05	0.03	0.04	0.04
SPAN Hu et al. (2020)	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
ImageForensicsOSN Wu et al. (2022)	0.20	0.09	0.18	0.09	0.30	0.08	0.23	0.07	0.09	0.05	0.20	0.08	0.13
MVSS-Net++ Dong et al. (2022)	0.06	0.03	0.09	0.05	0.15	0.03	0.12	0.03	0.07	0.04	0.10	0.04	0.07

Table 4: Evaluation of image forgery localization methods on the fully regenerated (FR) subsets of our TGIF2 dataset, for both the semantic (sem) and random (rand) versions (F1 score). The last rows show the IFL methods fine-tuned on the FR subsets. F1 scores above 0.7 are highlighted in bold.

IFL Method	Fine-tune set		SD2		SDXL		FLUX.1						Average
			SD2		SDXL		[schnell]		[dev]		Fill [dev]		Average
			Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	All
TruFor	All	Sem+Rand	0.49	0.39	0.68	0.83	0.87	0.94	0.82	0.88	0.76	0.85	0.72	0.78	0.75
		Sem	0.50	0.23	0.61	0.62	0.80	0.62	0.69	0.46	0.65	0.59	0.55	0.76	0.65
		Rand	0.31	0.40	0.50	0.80	0.69	0.92	0.62	0.85	0.63	0.82	0.55	0.76	0.58
	SD2	Sem	0.49	0.22	0.29	0.17	0.40	0.16	0.31	0.08	0.21	0.10	0.34	0.14	0.24

Table 5: Evaluation of individually fine-tuned image forgery localization methods on the fully regenerated (FR) subsets of our TGIF2 dataset, for both the semantic (Sem) and random (Rand) versions (F1 score). F1 scores above 0.7 are highlighted in a bold font. We notice that in-domain (i.e., same generative method) performance is relatively good, but generalization to out-of-domain (i.e., different generative method) performance is bad. Interestingly, when fine-tuning on only Sem masks, the evaluation on Rand subsets is lower than the corresponding Sem subsets, suggesting a potentially introduced bias towards semantics.

Panel of Figure 4 — Figure 4: An example of a fully regenerated inpainted image with SD2, using a bounding box of the tennis racket as mask (a, SD2 Sem), or a random mask in the image (d, SD2 Rand), along with the corresponding ground-truth masks (b, e) and the fine-tuned TruFor detection results (c, f). The fine-tuned TruFor was trained on the SD2-FR-Sem training set. The inpainted tennis racket in the SD2-FR-Sem test image is detected relatively well by the fine-tuned model. However, the inpainted random box of the SD2-FR-Rand subset is not detected; instead, the tennis racket is wrongly detected – suggesting that the fine-tuned model is be biased towards semantics or salient objects. Note that the inpainted random box, in fact, made notable changes to the sign in the background, and hence should be detectable by a better detection model.

Metric Class	Metric	SD2	PS	SDXL	FLUX.1
Metric Class	Metric	SD2	PS	SDXL	[schnell]	[dev]	Fill [dev]
Aesthetic Quality	NIMA mean $\uparrow$	5.05	5.16	4.88	5.16	5.28	5.21
Aesthetic Quality	GIQA $\uparrow$	-399	-373	-423	-383	-371	-424
Image-Text Matching	ITM $\uparrow$	0.28	0.32	0.26	0.31	0.30	0.25
Image-Text Matching	cos $\uparrow$	0.37	0.38	0.34	0.38	0.38	0.36
Preservation Fidelity (FR only)	PSNR $\uparrow$	27.1	N/A	18.1	32.5	32.5	28.5

Table 6: Average generative quality per (semantic) subset. In addition, the average detection performance is given as well (same as in Table 4 ). For each metric, we highlight the best performing subset in a bold font, and the worst performing subset in an italic font.

IFL Method	Version	Subsets	Aesthetic Quality		ITM		Preservation Fidelity
IFL Method	Version	Subsets	NIMA	GIQA	ITM	cos	PSNR	SSIM	LPIPS
TruFor	original	SP	0.36	0.42	-0.31	0.07	N/A	N/A	N/A
MMFusion	original	SP	0.30	0.39	-0.29	0.11	N/A	N/A	N/A
TruFor	fine-tuned	FR	0.28	0.34	-0.03	0.23	0.24	0.28	-0.32
MMFusion	fine-tuned	FR	0.28	0.35	-0.02	0.24	0.24	0.28	-0.32

Table 7: Spearman’s correlation and corresponding p-value between the IFL performance (F1 score), on the one hand, and the generative quality metrics, on the other. All correlation values are relatively low, with p-values equal to 0.000.

SID Method	Trained on	SD2		SDXL		FLUX.1						Average
		SD2		SDXL		[schnell]		[dev]		Fill [dev]		Average
		Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	All
CNNDetect Wang et al. (2020)	ProGAN	0.57	0.48	0.61	0.62	0.50	0.47	0.49	0.50	0.62	0.62	0.56	0.54	0.55
Dire Wang et al. (2023b)	ADM	0.49	0.52	0.62	0.64	0.49	0.54	0.48	0.52	0.48	0.49	0.51	0.54	0.53
FreqDetect Frank et al. (2020)	GANs	0.50	0.42	0.40	0.41	0.39	0.36	0.37	0.38	0.39	0.34	0.41	0.38	0.40
Fusing Ju et al. (2022)	ProGAN	0.55	0.48	0.47	0.47	0.55	0.53	0.55	0.56	0.62	0.62	0.55	0.53	0.54

Table 8: Evaluation of synthetic image detection methods on the semantic subsets of our TGIF2 Dataset (AUC Score). AUC scores above 0.8 are highlighted in a bold font.

IFL Method	Original	Super-resolution method
		Cubic interpolation		Real-ESRGAN
		x2	x4	x2	x4
CAT-Net Kwon et al. (2022)	0.87	0.76 -0.11	0.73 -0.14	0.06 -0.81	0.07 -0.80
TruFor Guillaro et al. (2023)	0.78	0.76 -0.02	0.76 -0.02	0.21 -0.57	0.22 -0.56
MMFusion Triaridis and Mezaris (2024)	0.69	0.65 -0.04	0.64 -0.05	0.18 -0.51	0.51 -0.51

Table 9: Evaluation of super-resolution attack on image forgery localization methods on spliced images of our TGIF2 Dataset, alongside a cubic interpolation attack as baseline (F1 score, along with the

\Delta

F1 Score).

SID Method	Trained on	Original	Super-resolution method
			Cubic interpolation		Real-ESRGAN
			x2	x4	x2	x4
DIMD Corvi et al. (2023)	LDM	0.88	0.87 –0.01	0.87 –0.01	0.75 –0.13	0.80 –0.08
UnivFD Ojha et al. (2023)	ProGAN	0.70	0.70 +0.00	0.69 –0.01	0.68 –0.02	0.68 –0.02
GramNet Liu et al. (2020)	GANs	0.76	0.76 +0.00	0.75 –0.01	0.50 –0.26	0.50 –0.26
LGrad Tan et al. (2023)	ProGAN	0.85	0.85 +0.00	0.84 –0.01	0.49 –0.36	0.48 –0.37

Table 10: Evaluation of super-resolution attack on synthetic image detection methods on fully regenerated images in our TGIF2 dataset, alongside a cubic interpolation attack as baseline (AUC Score, along with the

\Delta

AUC score).

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

IFL Method	SD2		PS	FLUX.1						Average
	SD2		PS	[schnell]		[dev]		Fill [dev]		Average
	Sem	Rand	Sem	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	All
PSCC-Net Liu et al. (2022)	0.12	0.13	0.31	0.02	0.06	0.01	0.03	0.06	0.17	0.10	0.10	0.10
SPAN Hu et al. (2020)	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.01	0.00	0.00	0.00
ImageForensicsOSN Wu et al. (2022)	0.16	0.05	0.25	0.22	0.07	0.16	0.04	0.08	0.07	0.18	0.06	0.12
MVSS-Net++ Dong et al. (2022)	0.05	0.02	0.05	0.11	0.04	0.08	0.02	0.06	0.07	0.07	0.04	0.06

Appendix

Table 11: Evaluation of image forgery localization methods on the spliced (SP) subsets of our TGIF2 dataset, for both the semantic (sem) and random (rand) versions ( IoU ). IoU values above 0.6 are highlighted in a bold font. Just as in Table 3 , some IFL methods (i.e., CAT-Net, TruFor & MMFusion) perform well on SP subsets.

IFL Method	SD2		SDXL		FLUX.1						Average
	SD2		SDXL		[schnell]		[dev]		Fill [dev]		Average
	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	All
PSCC-Net Liu et al. (2022)	0.03	0.02	0.03	0.05	0.01	0.01	0.01	0.01	0.02	0.03	0.02	0.03	0.02
SPAN Hu et al. (2020)	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
ImageForensicsOSN Wu et al. (2022)	0.14	0.05	0.12	0.05	0.22	0.05	0.16	0.04	0.06	0.03	0.14	0.04	0.09
MVSS-Net++ Dong et al. (2022)	0.04	0.02	0.06	0.03	0.11	0.02	0.09	0.02	0.05	0.03	0.07	0.02	0.05

Appendix

Table 12: Evaluation of image forgery localization methods on the fully regenerated (FR) subsets of our TGIF2 dataset, for both the semantic (sem) and random (rand) versions ( IoU ). The last rows show the IFL methods fine-tuned on All FR Sem subsets. Just as in Table 4 , the performance is low for all original IFL methods on FR subsets, while it is significantly higher for the fine-tuned variants.

IFL Method	Fine-tune set		SD2		SDXL		FLUX.1						Average
			SD2		SDXL		[schnell]		[dev]		Fill [dev]		Average
			Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	Sem	Rand	All
TruFor	All	Sem+Rand	0.37	0.28	0.58	0.75	0.80	0.90	0.74	0.82	0.67	0.78	0.63	0.71	0.67
		Sem	0.21	0.29	0.41	0.71	0.60	0.87	0.53	0.78	0.53	0.74	0.55	0.41	0.48
		Rand	0.38	0.15	0.50	0.52	0.71	0.52	0.59	0.37	0.55	0.50	0.46	0.68	0.57
	SD2	Sem	0.38	0.14	0.20	0.11	0.29	0.11	0.22	0.05	0.15	0.06	0.25	0.09	0.17

Appendix

Table 13: Evaluation of individually fine-tuned image forgery localization methods on the fully regenerated (FR) subsets of our TGIF2 dataset, for both the semantic (sem) and random (rand) versions ( IoU ). IoU values above 0.6 are highlighted in a bold font. Just as in Table 5 , we notice good in-domain performance but bad out-of-domain generalization, as well as a potential bias towards semantics.

Digital Image Forensics

Digital image forensics develops methods to detect, localize, explain, and trace image manipulation, synthetic content, and hidden data.

See all 112 papers

SEED: Simple ViT and Evolving Harness for Explainable Text Forgery Detection
Jun 19, 2026Kahim Wong, Kemou Li, Yiming Chen +2Image Forgery Detection Digital Image Forensics
State Key Laboratory of Internet of Things for Smart City, Department of Computer and Information Science, University of Macau · School of Information and Software Engineering, University of Electronic Science and Technology of China
RiSE: Residual Subspace Expert for Generalizable Text-Centric Image Forgery Localization
Oct 17, 2026Kahim Wong, Kemou Li, Yiming Chen +2Image Forgery Localization Text Forgery Detection
State Key Laboratory of Internet of Things for Smart City, University of Macau · School of Information and Software Engineering, University of Electronic Science and Technology of China
FUSED: Forensic-Semantic Mixture-of-Experts for AI Inpainting Detection and Localization
Aug 28, 2026Anton Nuzhdin, Marcel Worring, Ivona NajdenkoskaAI-Generated Image Detection Image Forgery Detection
Informatics Institute, University of Amsterdam Amsterdam, The Netherlands

Jun 19, 2026J/K move · Enter open · S save

SEED: Simple ViT and Evolving Harness for Explainable Text Forgery Detection

Kahim Wong, Kemou Li, Yiming Chen, Haiwei Wu+1

State Key Laboratory of Internet of Things for Smart City, Department of Computer and Information Science, University of Macau · School of Information and Software Engineering, University of Electronic Science and Technology of China

Image Forgery Detection Digital Image Forensics Digital Forensics Image Forgery Localization Text Forgery Detection

Open paper AskPDF arXiv

AI-assisted image editing threatens trust in financial, legal, and identity records. The GenText-Forensics Challenge at ACM MM 2026 addresses this by requiring structured forensic reports, in which integrating detection, pixel-level localization, and natural language explanation for multilingual text-centric forgery images. We present SEED, a modular system with three components. First, a similarity-guided pipeline augments training with diverse synthetic forgeries. Second, a single ViT, built on DINOv3 with LoRA adaptation, jointly performs detection and pixel-level localization while preserving pre-trained priors with minimal trainable parameters. Third, an evolving harness takes the detector's predictions and generates a complete forensic report via an MLLM, iteratively improved through a proposer-evaluator loop optimizing report quality. SEED ranked 3rd in the GenText-Forensics Challenge. Code and data are available at https://github.com/KahimWong/GenText-Forensics-3rd-Place.

Abstract

Digital Image Forensics

Explore similar work

SEED: Simple ViT and Evolving Harness for Explainable Text Forgery Detection

RiSE: Residual Subspace Expert for Generalizable Text-Centric Image Forgery Localization

FUSED: Forensic-Semantic Mixture-of-Experts for AI Inpainting Detection and Localization