Generative AI-Based Data Augmentation for Oral Lesion Classification: The PhotoMOCI Dataset and Benchmark

Marco Parola; Mario G. C. A. Cimino; Sabrina Senatore

cs.CVSep 28, 2026

Generative AI-Based Data Augmentation for Oral Lesion Classification: The PhotoMOCI Dataset and Benchmark

Authors: Marco Parola, Mario G. C. A. Cimino, Sabrina Senatore

Organizations: Aalborg University, Denmark. · University of Pisa, Italy. · University of Salerno, Italy.

Abstract

Early detection of oral cancer via photographic imaging presents a promising avenue for large-scale oral cavity screening. However, the development of robust deep learning models is frequently hampered by the scarcity of high-quality, annotated datasets. To address this limitation, a novel and well-curated resource, the Photographic Multi-purpose Oral Cancer Imaging (PhotoMOCI) dataset, is introduced for developing models across multiple diagnostic tasks in oral oncology. Then, a comprehensive benchmark study was conducted to investigate how various data augmentation strategies influence the performance of image classifiers. Our analysis spans different generative AI frameworks, evaluating the efficacy of traditional methods against advanced generative approaches, including Generative Adversarial Networks (GANs) and Diffusion Models (DMs). Additionally, we propose the Synthetic Image Filter (SIF), a mechanism to select specific samples based on two auxiliary models: Synthetic Proxy Classifier to ensure samples are representative of the target class and Synthetic Image Detector to verify they appear realistic, thereby selecting only the high-utility images that contribute to improving downstream performance. Across the evaluated datasets and classifiers, the best SIF-filtered setup improves accuracy over traditional augmentation in all cases, with gains of +1.73% and +2.35% on PhotoMOCI and +2.38% and +2.08% on KOCD for ResNet50 and ViT, respectively. Our findings reveal that while the direct application of generative data augmentation may yield performance drops, the integration of SIF, considering (i) how synthetic data looks real and (ii) how it reflects the discriminative features of the belonging class, provides a simple yet effective mechanism to filter out synthetic samples that confuse the classifier during training.

Figures & tables

Name	Year	#Imgs (w/ lesions)	Downstream tasks
Kaggle OC Lips and Tongue Shivam and Prakrut (2020)	2020	131 (87)	Binary image classification
Roboflow OC sagari vijay (2021)	2021	323 (156)	Multi-label image classification, Object detection, Semantic segmentation
Mendeley Oral Images Islam et al. (2023)	2023	373 (158)	Binary image classification
KOCD Rashid (2024)	2024	950 (500)	Binary image classification
PhotoMOCI (ours)	2026	700 (700)	Multi-label image classification, Object detection, Semantic segmentation, CBR

Table 1 : Existing photographic oral cancer datasets for computer vision. The column #Imgs (w/ lesions) reports total images and lesion‑positive ones (within round brackets).

Panel of Figure 1 — Figure 1 : PhotoMOCI metadata distribution by sex in (a) and age group in (b), stratified by pathology class.

Figure 2 : Overview of the proposed methodology for oral lesion classification. In the ochre dotted box, the Synthetic Image Generation based on generative models. In the green dotted box, the Synthetic Image Filter combining the Synthetic Proxy Classifier and the Synthetic Image Detector to select informative synthetic samples. In the light-blue dotted boxes, three training setups: using the original small dataset (top), the augmented synthetic dataset (middle), and the augmented synthetic filtered dataset (bottom), the latter aiming to improve generalization.

Panel of Figure 3 — Figure 3: Two examples of diffusion processes using SD(txt+img) with conditioning mechanisms based on text and image for neoplastic and traumatic lesions in (a) and (b) from PhotoMOCI, respectively. In each example, ten denoising processes with image conditioning performed at different steps. The red -bordered images indicate where the conditioning image is introduced. In the last row, with water-green borders, the original conditioning image injected at the 100th step, overriding earlier ones.

Synthetic img. gen. setup	FID $\downarrow$	KID $\downarrow$	Prec. $\uparrow$	Recall $\uparrow$	Cov. $\uparrow$
Tr. aug.	230.30	0.126	0.317	0.183	0.058
SD(txt)	223.92	0.079	0.149	0.660	0.191
SD(txt+img)	179.24	0.072	0.543	0.319	0.521
SG3(lbl)	195.77	0.099	0.404	0.234	0.287
AC-SG3	290.42	0.230	0.426	0.011	0.032

Table 2 : Synthetic Image Generation results on PhotoMOCI.

Synthetic img. gen. setup	FID $\downarrow$	KID $\downarrow$	Prec. $\uparrow$	Recall $\uparrow$	Cov. $\uparrow$
Tr. aug.	262.91	0.126	0.284	0.632	0.242
SD(txt)	225.38	0.074	0.273	0.762	0.231
SD(txt+img)	143.87	0.034	0.867	0.573	0.888
SG3(lbl)	169.57	0.062	0.888	0.112	0.748
AC-SG3	319.78	0.245	0.035	0.035	0.042

Table 3 : Synthetic Image Generation results on KOCD.

Panel of Figure 4 — Figure 4 : Synthetic image examples generated from the PhotoMOCI dataset using four synthetic image generation setups: Stable Diffusion conditioned on text (top-left), Stable Diffusion conditioned on text and image (top-right), StyleGAN3 conditioned on text (bottom-left), and AC-SG3 (bottom-right). Within each block, the first, second, and third rows correspond to traumatic, aphthous, and neoplastic lesions, respectively.

Panel of Figure 5 — Figure 5 : Synthetic image examples generated from the KOCD using four synthetic image generation setups: Stable Diffusion conditioned on text (top-left), Stable Diffusion conditioned on text and image (top-right), StyleGAN3 conditioned on text (bottom-left), and AC-SG3 (bottom-right). Within each block, the first row corresponds to the cancer class, and the second row corresponds to the non-cancer class.

Synthetic img. gen. setup	Single image time (s)	PhotoMOCI 2100 images	KOCD 3000 images
SD(txt)	1.86 $\pm$ 0.05	1 h 05 min 06 s	1 h 33 min 00 s
SD(txt+img)	0.82 $\pm$ 0.06	28 min 42 s	41 min 00 s
SG3(lbl)	0.045 $\pm$ 0.003	1 min 35 s	2 min 15 s
AC-SG3	0.048 $\pm$ 0.004	1 min 41 s	2 min 24 s

Table 4 : Computational time for generating synthetic images on an NVIDIA A100 GPU.

Dataset	Synthetic img. gen. setup	Accuracy
Dataset	Synthetic img. gen. setup	SPC $\uparrow$	SID $\downarrow$
PhotoMOCI (ours)	SD(txt)	0.368 $\pm$ 0.045	0.788 $\pm$ 0.032
	SD(txt+img)	0.771 $\pm$ 0.037	0.643 $\pm$ 0.040
	SG3(lbl)	0.740 $\pm$ 0.029	0.596 $\pm$ 0.038
	AC-SG3	0.893 $\pm$ 0.036	0.921 $\pm$ 0.036
KOCD Rashid (2024)	SD(txt)	0.617 $\pm$ 0.025	0.709 $\pm$ 0.030

Table 5 : SIF modules performance: classification results of SPC and SID on PhotoMOCI and KOCD considering the four synthetic data generation.

Synthetic img. gen. setup	PhotoMOCI				KOCD
Synthetic img. gen. setup	Total	Traumatic	Aphthous	Neoplastic	Total	Cancer	Non-cancer
SD(txt)	139 (6.6%)	30 (4.3%)	83 (11.9%)	26 (3.7%)	308 (10.3%)	91 (6.1%)	217 (14.5%)
SD(txt+img)	470 (22.4%)	157 (22.4%)	119 (17.0%)	194 (27.7%)	653 (21.8%)	362 (24.1%)	291 (19.4%)
SG3(lbl)	367 (17.5%)	141 (20.1%)	137 (19.6%)	89 (12.7%)	776 (25.9%)	415 (27.7%)	361 (24.1%)
AC-SG3	70 (3.3%)	12 (1.7%)	19 (2.7%)	39 (5.6%)	152 (5.1%)	24 (1.6%)	128 (8.5%)

Table 6 : Number and percentage of synthetic images retained by SIF on PhotoMOCI and KOCD, reported in aggregate and by class.

Model	Conditioning	SIF	Acc	Prec	Recall	F1-score
ResNet He et al. (2016)	No aug.	✗	0.81050 $\pm$ 0.00532	0.81538 $\pm$ 0.00553	0.81076 $\pm$ 0.00581	0.81414 $\pm$ 0.00601
	Tr.aug.(base)	✗	0.82519 $\pm$ 0.00600	0.83130 $\pm$ 0.00572	0.82316 $\pm$ 0.00603	0.82873 $\pm$ 0.00612
	SD(txt)		0.74153 $\pm$ 0.00590	0.74797 $\pm$ 0.00619	0.74225 $\pm$ 0.00528	0.74550 $\pm$ 0.00493
	SD(txt)	✓	0.80076 $\pm$ 0.00544	0.82653 $\pm$ 0.00485	0.80567 $\pm$ 0.00521	0.81237 $\pm$ 0.00495
	SD(txt+img)		0.83655 $\pm$ 0.00520	0.83704 $\pm$ 0.00446	0.83679 $\pm$ 0.00499	0.83698 $\pm$ 0.00524
	SD(txt+img)	✓	0.84253 $\pm$ 0.00498	0.85137 $\pm$ 0.00507	0.84294 $\pm$ 0.00481	0.85095 $\pm$ 0.00511

Table 7 : Classification performance on PhotoMOCI using ResNet and ViT under different data augmentation strategies. The evaluated setups include training without augmentation (No aug.) and with traditional augmentation (Tr. aug.), marked with ✗ since SIF is not applicable, as well as four generative-AI–based setups evaluated both without and with SIF, distinguished by ✓. Highlighted in gray , the setups where SIF improves performance over non-SIF counterparts and surpasses traditional augmentation. Highlighted in bold , the best-performing augmentation setup for each classifier.

Model	Conditioning	SIF	Acc	Prec	Recall	F1-score
ResNet He et al. (2016)	No aug.	✗	0.90935 $\pm$ 0.00453	0.90028 $\pm$ 0.00452	0.90074 $\pm$ 0.00449	0.90026 $\pm$ 0.00454
	Tr.aug.(base)	✗	0.93109 $\pm$ 0.00554	0.93103 $\pm$ 0.00547	0.93120 $\pm$ 0.00563	0.93095 $\pm$ 0.00576
	SD(txt)		0.88564 $\pm$ 0.00637	0.89160 $\pm$ 0.00656	0.88713 $\pm$ 0.00625	0.89121 $\pm$ 0.00700
	SD(txt)	✓	0.91616 $\pm$ 0.00410	0.91937 $\pm$ 0.00474	0.91668 $\pm$ 0.00422	0.91871 $\pm$ 0.00390
	SD(txt+img)		0.95011 $\pm$ 0.00498	0.95225 $\pm$ 0.00501	0.94917 $\pm$ 0.00533	0.95075 $\pm$ 0.00511
	SD(txt+img)	✓	0.95492 $\pm$ 0.00467	0.95719 $\pm$ 0.00501	0.95266 $\pm$ 0.00461	0.95543 $\pm$ 0.00521

Table 8 : Classification performance on KOCD using ResNet and ViT under different data augmentation strategies. The evaluated setups include training without augmentation (No aug.) and with traditional augmentation (Tr. aug.), marked with ✗ since SIF is not applicable, as well as four generative-AI–based setups evaluated both without and with SIF, distinguished by ✓. Highlighted in gray , the setups where SIF improves performance over non-SIF counterparts and surpasses traditional augmentation. Highlighted in bold , the best-performing augmentation setup for each classifier.

Panel of Figure 6 — Figure 6 : Confusion matrices for PhotoMOCI comparing traditional augmentation and SD(txt+img)+SIF for ResNet50 and ViT.

Dataset	Model	$\boldsymbol{\chi^{2}}$	$\boldsymbol{p}$	Kendall’s $\boldsymbol{W}$
KOCD	ResNet	54.938	$<0.0001$	0.610
KOCD	ViT	53.520	$<0.0001$	0.595
PhotoMOCI	ResNet	56.749	$<0.0001$	0.631
PhotoMOCI	ViT	50.787	$<0.0001$	0.564

Table 9 : Overall statistical significance analysis using Friedman tests on accuracy.

	KOCD				PhotoMOCI
Synthetic img. gen. setup	ResNet		ViT		ResNet		ViT
	$\Delta$ Acc.	$p$	$\Delta$ Acc.	$p$	$\Delta$ Acc.	$p$	$\Delta$ Acc.	$p$
SD(txt)+SIF	-1.49	0.0645	-0.59	0.3750	-2.44	0.0371	-0.84	0.4316
SD(txt+img)+SIF	+2.38	0.0371	+2.08	0.0059	+1.73	0.0273	+2.35	0.0195
SG3(lbl)+SIF	+1.72	0.0488	+1.17	0.0934	+1.08	0.0754	+1.79	0.0515
AC-SG3+SIF	+0.44	0.6953	+0.21	0.6250	-1.53	0.0195	-0.17	0.6250

Table 10 : Planned Wilcoxon comparisons between synthetic+SIF setups and traditional augmentation on accuracy. Values report accuracy difference in percentage points and

p

-value.

	KOCD				PhotoMOCI
Synthetic img. gen. setup	ResNet		ViT		ResNet		ViT
	$\Delta$ Acc.	$p$	$\Delta$ Acc.	$p$	$\Delta$ Acc.	$p$	$\Delta$ Acc.	$p$
SD(txt)	+3.05	0.0039	+3.58	0.0020	+5.92	0.0020	+6.31	0.0020
SD(txt+img)	+0.48	0.0566	+0.71	0.0316	+0.60	0.0566	+0.80	0.0316
SG3(lbl)	+0.81	0.0754	+0.63	0.0750	+0.32	0.0250	+0.57	0.0695
AC-SG3	+0.74	0.0602	+0.91	0.0223	+1.84	0.0195	+1.33	0.0324

Table 11 : Planned Wilcoxon comparisons between SIF-filtered and unfiltered synthetic augmentation on accuracy. Values report accuracy difference in percentage points and

p

-value.

Panel of Figure 7 — Figure 7 : Ablation study evaluating the impact of SPC and SID within the SIF design on (a) PhotoMOCI and (b) KOCD using ViT. The study compares training without augmentation (blue bar) to training with integrated synthetic data: unfiltered (ochre bar), filtered using SPC only (water-green bar), SID only (orange bar), and SIF (pink bar).

Panel of Figure 8 — Figure 8 : SID threshold calibration on (a) PhotoMOCI and (b) KOCD. Each plot reports accuracy as a function of the SID threshold $\tau$ for ResNet50 (left) and ViT (right), with shaded 95% confidence intervals over 10 runs and a marker indicating the best threshold.

Explore similar work

Cross-Domain Adversarial Augmentation: Stabilizing GANs for Medical and Handwriting Data Scarcity
May 3, 2026Md. Sohanuzzaman Soad, Mahady Al Hady, S M Rafiuddin Rifat +1Data Augmentation Generative Adversarial Network
Department of Computer Science and Engineering University of Asia Pacific Dhaka, Bangladesh
Do Synthetic Brain MRIs Reliably Improve Tumour Classification? A StyleGAN2-ADA Class-Plane Augmentation Study on BRISC 2025
May 21, 2026José Rafael Noriega CedeñoMedical Image Generation Magnetic Resonance Imaging
Faculty of Applied Sciences and Technology
Searching for Robust Augmentations to Improve Out-of-Domain Generalization in Dermoscopic Skin Cancer Classification
Jul 29, 2026Alexander Kozachok, Ilya Latyshev, Evgeny Karpulevich +3Skin Lesion Classification Data Augmentation
Trusted AI Research Center, Russian Academy of Sciences, 109004 Moscow, Russia

May 3, 2026, cs.CVJ/K move · Enter open · S save

Cross-Domain Adversarial Augmentation: Stabilizing GANs for Medical and Handwriting Data Scarcity

Md. Sohanuzzaman Soad, Mahady Al Hady, S M Rafiuddin Rifat, Sudip Ghose

Department of Computer Science and Engineering University of Asia Pacific Dhaka, Bangladesh

Data Augmentation Generative Adversarial Network Medical Image Generation Medical Images Handwriting

Open paper AskPDF arXiv

Generative Adversarial Networks (GANs) can help overcome data scarcity in computer vision tasks by generating additional training samples. In this work, we explore generative data augmentation in two low-resource domains: Bangla handwritten character recognition and chest X-ray image analysis. We use DCGAN-based models trained on 64x64 images to generate synthetic samples and evaluate their quality using Inception Score (IS), Fréchet Inception Distance (FID), and visualization methods such as t-SNE and UMAP. To measure practical usefulness, we train image classifiers using real data and a combination of real and synthetic data. Experimental results show that synthetic augmentation improves data diversity and consistently increases classification performance in limited-data settings. We also investigate training stability techniques, including gradient penalty and spectral normalization, and perform ablation studies on synthetic-to-real data ratios and sample filtering strategies. In addition, we discuss challenges related to medical image evaluation, dataset licensing, and privacy concerns of synthetic data. Our approach is simple, reproducible, and provides a strong baseline for generative augmentation in resource-constrained imaging applications.

Abstract

Explore similar work

Cross-Domain Adversarial Augmentation: Stabilizing GANs for Medical and Handwriting Data Scarcity

Do Synthetic Brain MRIs Reliably Improve Tumour Classification? A StyleGAN2-ADA Class-Plane Augmentation Study on BRISC 2025

Searching for Robust Augmentations to Improve Out-of-Domain Generalization in Dermoscopic Skin Cancer Classification