Operative Image Spaces

Background The validation of synthetic dermatological images generated by Generative Adversarial Networks (GANs) [1] is crucial for their integration into clinical and research workflows. Despite rapid progress in image synthesis, a standardized framework for evaluating the realism and diagnostic utility of synthetic skin lesions through expert review is still lacking [2]. Existing automated evaluation metrics, while informative, do not always align with human perception and diagnostic expectations. Particularly in medical domains, subtle visual cues and contextual interpretation often elude algorithmic assessment [3]. Human evaluations remain the most direct means of determining whether synthetic images capture the nuanced features necessary for clinical utility. Without structured expert-based validation, synthetic images may introduce bias or mislead models and clinicians, hampering their responsible deployment in diagnostic support systems, training datasets, or educational tools. Objectives This study aims to conduct an expert-based qualitative evaluation of synthetic melanoma images. Specifically, it investigates the subjective perception of image realism, diagnostic quality, and the recognizability of key dermoscopic features. By engaging dermatologists in a blinded assessment of synthetic and real images, we seek to establish a foundation for systematically validating synthetic dermatological data for use in AI development, medical education, and clinical decision support. This work emphasizes the importance of subjective expert validation as a complement to technical performance metrics in assessing the fidelity of GAN-generated skin lesion images. Materials and Methods StyleGAN3-T [4] was trained on a dataset of dermoscopic images of melanoma [5–7] with adaptive discriminator augmentation and transfer learning. A total of 25 synthetic melanoma images were generated and randomly mixed with 25 real melanoma images, resulting in a 50-image dataset. Seventeen board-certified dermatologists with varying levels of experience (low <4 years, medium 5–8 years, high >8 years) participated in the evaluation. Participants were blinded to image origin and asked to classify each image as real or synthetic. They also assessed the presence of 16 defined dermoscopic patterns according to standardized definitions and rated four dimensions—image quality, skin texture, visual realism, and color realism—on a 7-point Likert scale. Additionally, participants reported their confidence in each classification decision. Statistical analyses included Chi-square tests for categorical comparisons, and Fleiss’ Kappa and Krippendorff’s Alpha were used to measure inter-rater agreement. Results Real images were consistently rated higher than synthetic images across all qualitative dimensions: image quality (high: 15.8% real vs. 11.3% synthetic), skin texture (high: 22.4% vs. 13.4%), and visual realism (high: 22.6% vs. 13.2%), all with p < 0.001. Confidence in evaluations was also significantly greater for real images, with high confidence reported in 17.4% of real cases compared to 8.7% for synthetic ones (p < 0.001).Regarding the recognition of image origin, the overall classification accuracy was 64%. Real images were correctly identified in 73% of cases, while only 56% of synthetic images were correctly classified as synthetic. Accuracy increased with expertise: from 59% in the low-experience group to 71% among high-experience dermatologists. Similarly, higher self-reported confidence was associated with improved performance (accuracy 74% at high confidence level). Recognition of specific dermoscopic features showed differences between real and synthetic images. The blue-white veil was detected in 29.1% of real images compared to 13.8% of synthetic ones (p < 0.001), and shiny white streaks in 22.6% vs. 7.9% (p < 0.001). Conversely, synthetic images were more frequently associated with irregular pigmented blotches (45.0% vs. 30.9%, p < 0.001). The multicomponent pattern, typically indicative of melanoma complexity, was identified in 40.6% of real images versus only 23.2% of synthetic ones (p < 0.001), suggesting a gap in the synthetic images’ structural fidelity (Table 1). Inter-rater agreement for the classification of real versus synthetic images was low, with a Fleiss’ kappa of 0.183. Pattern recognition agreement also remained weak (e.g., kappa < 0.3 for most features), underscoring variability in expert interpretations. Further subgroup analyses showed that images rated as highly realistic or evaluated with high confidence were more likely to be classified correctly, with accuracy rising to 74% in the highest-confidence subgroup. Conclusions Synthetic melanoma lesions generated using StyleGAN3-T demonstrate visually convincing features and were frequently perceived as real, yet consistently underperformed compared to real images in diagnostic quality and structural detail. Participants often struggled to distinguish synthetic from real lesions, particularly when realism ratings were medium to high. Critical diagnostic patterns, such as the blue-white veil and shiny white streaks, were significantly underrepresented in synthetic images. These limitations were reflected in the lower classification confidence and weaker inter-rater agreement. Despite these challenges, the study highlights the potential of synthetic data to approach realism levels sufficient for research and educational use. Qualitative validation by dermatologists is essential to benchmark the readiness of synthetic images for real-world medical applications. As generative models continue to evolve, expert evaluation should remain a key component of validation pipelines to ensure clinical and pedagogical reliability.

Are synthetic cytology images ready for prime time? A comparative assessment of real and synthetic urine cytology images.

This paper presents a systematic method to create universally applicable synthetic training image datasets for computer vision-based construction object detection. The synthetic images created by inserting a virtual object of interest into a real site image allows us to minimize the time and effort for training image data collection and annotation. In addition, the use of synthetic images has an additional benefit that training images can be easily customized for a target construction site by considering the context of the site (e.g., different background scenes, camera positions, and angles) and the possible variability of target objects to be detected (e.g., different sizes, locations, rotation angles, and postures) on images. An automated approach proposed in this study attempts to systematically create the synthetic images using the Unity game engine in which context- and variability-related parameters can be controlled. The proposed method was validated by training a deep learning-based object detection algorithm [i.e., a faster regions with convolutional neural network (R-CNN) model] with synthetic images and testing it on real images from earthwork construction sites to detect an excavator. The CNN models trained with synthetic images showed an average precision value of more than 90%; in particular, the classifier using synthetic images outperformed the one using real site images. The detection results also demonstrated an improved capability to capture the high irregularity of a construction object on images when using techniques of context customization and variability randomization. The findings from this study demonstrate the feasibility and practicality of the use of synthetic images for vision-based approaches in a construction domain. Ultimately, the proposed approach serves as an alternative way to build comprehensive image datasets for construction entities, contributing to facilitating vision-based studies on construction.

Deep learning (DL) used for discriminative tasks in ophthalmology, such as diagnosing diabetic retinopathy or age-related macular degeneration (AMD), requires large image data sets graded by human experts to train deep convolutional neural networks (DCNNs). In contrast, generative DL techniques could synthesize large new data sets of artificial retina images with different stages of AMD. Such images could enhance existing data sets of common and rare ophthalmic diseases without concern for personally identifying information to assist medical education of students, residents, and retinal specialists, as well as for training new DL diagnostic models for which extensive data sets from large clinical trials of expertly graded images may not exist. To develop DL techniques for synthesizing high-resolution realistic fundus images serving as proxy data sets for use by retinal specialists and DL machines. Generative adversarial networks were trained on 133 821 color fundus images from 4613 study participants from the Age-Related Eye Disease Study (AREDS), generating synthetic fundus images with and without AMD. We compared retinal specialists' ability to diagnose AMD on both real and synthetic images, asking them to assess image gradability and testing their ability to discern real from synthetic images. The performance of AMD diagnostic DCNNs (referable vs not referable AMD) trained on either all-real vs all-synthetic data sets was compared. Accuracy of 2 retinal specialists (T.Y.A.L. and K.D.P.) for diagnosing and distinguishing AMD on real vs synthetic images and diagnostic performance (area under the curve) of DL algorithms trained on synthetic vs real images. The diagnostic accuracy of 2 retinal specialists on real vs synthetic images was similar. The accuracy of diagnosis as referable vs nonreferable AMD compared with certified human graders for retinal specialist 1 was 84.54% (error margin, 4.06%) on real images vs 84.12% (error margin, 4.16%) on synthetic images and for retinal specialist 2 was 89.47% (error margin, 3.45%) on real images vs 89.19% (error margin, 3.54%) on synthetic images. Retinal specialists could not distinguish real from synthetic images, with an accuracy of 59.50% (error margin, 3.93%) for retinal specialist 1 and 53.67% (error margin, 3.99%) for retinal specialist 2. The DCNNs trained on real data showed an area under the curve of 0.9706 (error margin, 0.0029), and those trained on synthetic data showed an area under the curve of 0.9235 (error margin, 0.0045). Deep learning-synthesized images appeared to be realistic to retinal specialists, and DCNNs achieved diagnostic performance on synthetic data close to that for real images, suggesting that DL generative techniques hold promise for training humans and machines.

This work classifies color images of ships attained using cameras mounted on ships and in harbors. Our data-sets contain 9 different types of ship with 18 different perspectives for our training set, development set and testing set. The training data-set contains modeled synthetic images; development and testing data-sets contain real images. The database of real images was gathered from the internet, and 3D models for synthetic images were imported from Google 3D Warehouse. A key goal in this work is to use synthetic images to increase overall classification accuracy. We present a novel approach for autonomous segmentation and feature extraction for this problem. Support vector machine is used for multi-class classification. This work reports three experimental results for multi-class ship classification problem. First experiment trains on a synthetic image data-set and tests on a real image data-set, and obtained accuracy is 87.8%. Second experiment trains on a real image data-set and tests on a separate real image data-set, and obtained accuracy is 87.8%. Last experiment trains on real + synthetic image data-sets (combined data-set) and tests on a separate real image data-set, and obtained accuracy is 93.3%.

This research investigates the usefulness and efficacy of synthetic ruler images for the development of a deep learning-based ruler detection algorithm. Synthetic images offer a compelling alternative to real-world images as data sources in the development and advancement of computer vision systems. This research aims to answer whether using a synthetic dataset of ruler images is sufficient for training an effective ruler detector and to what extent such a detector could benefit from including synthetic images as a data source. The article presents the procedural method for generating synthetic ruler images, describes the methodology for evaluating the synthetic dataset using trained convolutional neural network (CNN)-based ruler detectors, and shares the compiled synthetic and real ruler image datasets. It was found that the synthetic dataset yielded superior results in training the ruler detectors compared with the real image dataset. The results support the utility of synthetic datasets as a viable and advantageous approach to training deep learning models, especially when real-world data collection presents significant logistical challenges. The evidence presented here strongly supports the idea that when carefully generated and used, synthetic data can effectively replace real images in the development of CNN-based detection systems.

Lunar exploration through both robotic and planned human controlled missions are being pursued through a variety of studies and missions from ESA, NASA and other space agencies. These missions are increasingly defining stringent requirements for guidance of approach, surface relative navigation, and hazard detection, for pin-point landings on difficult terrain under contrasting illumination conditions. So autonomous GNC and hazard avoidance systems are continuing to be developed and validated to meet mission requirements. This paper describes the creation of two new lunar simulations to cover both a long-distance approach over a hemisphere of the Moon, and a pinpoint landing, utilising the highest quality available Digital Elevation Models (DEMs). These are imported into the Planet and Asteroid Natural Scene Generation Utility (PANGU) tool [1], enhanced with representative terrain and small-scale features below the viable resolution of the DEMs, and high-frequency texture to obtain a large, high-resolution 3D model. Synthetic surface images are rendered with appropriate lighting and camera- distortion in real-time. These scenarios are not tailored for specific missions but are instead designed to be representative of potential missions which can be modified with different input data and generation parameters as required. While there is a large dataset of real images of lunar terrain, these are insufficient for training, testing and evaluating GNC and hazard avoidance systems. Synthetic image generation is one approach to augment the training, testing and evaluating autonomous navigation and landing systems, where the synthetic images can be generated with sufficient realism. Synthetic images can also be useful for training deep learning vision components of GNC systems and for closed-loop test with framerates to match the real-time requirements. However, generating realistic synthetic images in real-time to support the approach and landing phases is a challenging task due to the massive models required and an expected image generation frame rate of around 10 Hz. The PANGU 3D models are based on the best available DEMs of the Moon, which are data products from current and previous missions such as Lunar Reconnaissance Orbiter (LRO), Kaguya and the Chang’e series of missions. The highest quality available lunar DEMs vary in range and resolution from global elevation datasets such as the combined LRO/Kaguya SLDEM [2] with a maximum resolution of ~60m per pixel at the equator, to higher resolution DEM products of local sections of around 5m per pixel from the Lunar Orbiter Laser Altimeter (LOLA) instrument [3], and DEMs derived from stereo images from the LRO Camera (LROC) and Narrow Angle Camera (NAC). The first scenario is generated from the SLDEM DEM to simulate a low altitude descent trajectory with a swath strip covering half a lunar circumference, followed by a simulated manoeuvre to descend towards a target landing site. The SLDEM is imported into PANGU and converted to floating-point PDS format for further processing. The resolution of the section of the model along the descent trajectory is increased with representative fractal terrain by multiple factors of two towards the target landing site, to avoid clear resolution boundaries. Distributions of craters and boulders, based on size-density values from literature for the terrain type and region, are generated to seamlessly embed small scale features to the model below a resolution not clearly defined in the imported base DEM. Camera model parameters are specified to simulate noise and distortion for increased realism. The PANGU model is generated, and a sequence of images of it are rendered with appropriate lighting and shadow effects with the images encoded into a demonstration video using the FFmpeg tool. A second scenario is generated from a high-resolution DEM to simulate the final approach and pinpoint landing to a pre-selected site on challenging terrain near the lunar South Pole. Recently improved LOLA DEMs of selected terrain sections in the South Pole Region are now available, free from significant artefacts, with a horizontal resolution around 5m [3]. A representative DEM is selected with a primary landing site on a sunlit peak surrounded by deep shadows as a suitable South Pole landing site. Multiple resolution enhancement regions are defined surrounding other potential landing sites. A diameter density distribution of small craters less than 10m in diameter (which aren’t defined in the 5m DEM) is specified and used to generate a realistically representative list of craters to add to the multi-resolution terrain model as appropriate to each higher resolution layer. As boulders are a potential significant hazard for any lander, multiple boulder forms are specified, and a boulder size-density distribution is taken from literature data to generate a list of boulders with varying shapes, to add to the terrain model around the target landing sites. High-frequency texture is also added to represent surface roughness below the resolution of the model. A PANGU 3D model is generated from which multiple image sequences are rendered to simulate landing on the different target sites. The full paper will include full details of the imported DEMs and the PANGU models used to generate the images for these representative scenarios. They will also show comparisons between the real and synthetic images to demonstrate the realism and validity of the simulated images, and include the full details of resolution, size and frame rates obtained of the simulated descents. PANGU was developed by the University of Dundee for ESA and is being used on many European activities aimed at producing precise, robust planetary lander and rover guidance systems. [1] Martin, I., Dunstan, M., Parkes, S., & Gestido, M. S., “Testing Vision-based Guidance and Navigation Systems for Entry Descent and Landing Operations”. In IAC 2018 Conference Proceedings (pp. 1-9), 2018, [IAC-18,D1,3,4,x42780]. [2] LRO LOLA Digital Elevation Model Co-registered with Selene Data (SLDEM), website: https://ode.rsl.wustl.edu/moon/pagehelp/Content/Missions_Instruments/LRO/LOLA/SLDEM.htm. [3] Barker, M. K., Mazarico, E. M. and Restrepo C. I., (NASA GSFC), “Topographic Models from the Lunar Orbiter Laser Altimeter (LOLA) in Support of Terrain Relative Navigation at the Moon”, Paper SIW22–23, 3rd Space Imaging Workshop, Atlanta, USA, Oct 2023.

To realize the full potential of deep learning for medical imaging, large annotated datasets are required for training. Such datasets are difficult to acquire due to privacy issues, lack of experts available for annotation, underrepresentation of rare conditions, and poor standardization. The lack of annotated data has been addressed in conventional vision applications using synthetic images refined via unsupervised adversarial training to look like real images. However, this approach is difficult to extend to general medical imaging because of the complex and diverse set of features found in real human tissues. We propose a novel framework that uses a reverse flow, where adversarial training is used to make real medical images more like synthetic images, and clinically-relevant features are preserved via self-regularization. These domain-adapted synthetic-like images can then be accurately interpreted by networks trained on large datasets of synthetic medical images. We implement this approach on the notoriously difficult task of depth-estimation from monocular endoscopy which has a variety of applications in colonoscopy, robotic surgery, and invasive endoscopic procedures. We train a depth estimator on a large data set of synthetic images generated using an accurate forward model of an endoscope and an anatomically-realistic colon. Our analysis demonstrates that the structural similarity of endoscopy depth estimation in a real pig colon predicted from a network trained solely on synthetic data improved by 78.7% by using reverse domain adaptation.

In recent years, considerable work has been conducted on the development of synthetic medical images, but there are no satisfactory methods for evaluating their medical suitability. Existing methods mainly evaluate the quality of noise in the images, and the similarity of the images to the real images used to generate them. For this purpose, they use feature maps of images extracted in different ways or distribution of images set. Then, the proximity of synthetic images to the real set is evaluated using different distance metrics. However, it is not possible to determine whether only one synthetic image was generated repeatedly, or whether the synthetic set exactly repeats the training set. In addition, most evolution metrics take a lot of time to calculate. Taking these issues into account, we have proposed a method that can quantitatively and qualitatively evaluate synthetic images. This method is a combination of two methods, namely, FMD and CNN-based evaluation methods. The estimation methods were compared with the FID method, and it was found that the FMD method has a great advantage in terms of speed, while the CNN method has the ability to estimate more accurately. To evaluate the reliability of the methods, a dataset of different real images was checked.

Deep learning can bring time savings and increased reproducibility to medical image analysis. However, acquiring training data is challenging due to the time-intensive nature of labeling and high inter-observer variability in annotations. Rather than labeling images, in this work we propose an alternative pipeline where images are generated from existing high-quality annotations using generative adversarial networks (GANs). Annotations are derived automatically from previously built anatomical models and are transformed into realistic synthetic ultrasound images with paired labels using a CycleGAN. We demonstrate the pipeline by generating synthetic 2D echocardiography images to compare with existing deep learning ultrasound segmentation datasets. A convolutional neural network is trained to segment the left ventricle and left atrium using only synthetic images. Networks trained with synthetic images were extensively tested on four different unseen datasets of real images with median Dice scores of 91, 90, 88, and 87 for left ventricle segmentation. These results match or are better than inter-observer results measured on real ultrasound datasets and are comparable to a network trained on a separate set of real images. Results demonstrate the images produced can effectively be used in place of real data for training. The proposed pipeline opens the door for automatic generation of training data for many tasks in medical imaging as the same process can be applied to other segmentation or landmark detection tasks in any modality. The source code and anatomical models are available to other researchers.11https://adgilbert.github.io/data-generation/

This study explores the use of deep generative models to create synthetic ultrasound images for the detection of hemarthrosis in hemophilia patients. Addressing the challenge of sparse datasets in rare disease diagnostics, the study aims to enhance AI model robustness and accuracy through the integration of domain knowledge into the synthetic image generation process. The study employed two ultrasound datasets: a base dataset (Db) of knee recess distension images from non-hemophiliac patients and a target dataset (Dt) of hemarthrosis images from hemophiliac patients. The synthetic generation framework included a content generator (Gc) trained on Db and a context generator (Gs) to adapt these images to match Dt's context. This approach generated a synthetic target dataset (Ds), primed for AI training in rare disease research. The assessment of synthetic image generation involved expert evaluations, statistical analysis, and the use of domain-invariant perceptual distance and Fréchet inception distance for quality measurement. Expert evaluation revealed that images produced by our synthetic generation framework were comparable to real ones, with no significant difference in overall quality or anatomical accuracy. Additionally, the use of synthetic data in training convolutional neural networks demonstrated robustness in detecting hemarthrosis, especially with limited sample sizes. This study presents a novel approach for generating synthetic ultrasound images for rare disease detection, such as hemarthrosis in hemophiliac knees. By leveraging deep generative models and integrating domain knowledge, the proposed framework successfully addresses the limitations of sparse datasets and enhances AI model training and robustness. The synthetic images produced are of high quality and contribute significantly to AI-driven diagnostics in rare diseases, highlighting the potential of synthetic data in medical imaging.

Deep learning approaches require enough training samples to perform well, but it is a challenge to collect enough real training data and label them manually. In this letter, we propose a practical framework for automatically generating content-rich synthetic images with ground-truth annotations. By rendering 3-D CAD models, we generate two synthetic aircraft image data sets with wide distribution (Syn N and Syn U). For improving the quality of synthetic images, we propose a multiscale attention module which enhances the Cycle-Consistent Adversarial Network (CycleGAN) in spatial and channel dimensions. Then, we compare the synthetic images before and after translation qualitatively and quantitatively. Experiments on Northwestern Polytechnical University (NWPU) very high resolution (VHR)-10, University of Chinese Academy of Sciences, orientation robust object detection in aerial images (UCAS-AOD), and benchmark for object DetectIon in Optical Remote sensing images (DIOR) data sets demonstrate that synthetic data augmentation can improve the performance of aircraft detection in remote sensing images, especially when real data are insufficient. Synthetic data are available at: <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://weix-liu.github.io/</uri> .

This paper investigates how the size of virtual space and objects within it affect the threshold range of relative translation gains, a Redirected Walking (RDW) technique that scales the user&#x2019;s movement in virtual space in different ratios for the width and depth. While previous studies assert that a virtual room&#x2019;s size affects relative translation gain thresholds on account of the virtual horizon&#x2019;s location, additional research is needed to explore this assumption through a structured approach to visual perception in Virtual Reality (VR). We estimate the relative translation gain thresholds in six spatial conditions configured by three room sizes and the presence of virtual objects (3 &#x00D7; 2), which were set according to differing Angles of Declination (AoDs) between eye-gaze and the forward-gaze. Results show that both size and virtual objects significantly affect the threshold range, it being greater in the large-sized condition and furnished condition. This indicates that the effect of relative translation gains can be further increased by constructing a perceived virtual movable space that is even larger than the adjusted virtual movable space and placing objects in it. Our study can be applied to adjust virtual spaces in synchronizing heterogeneous spaces without coordinate distortion where real and virtual objects can be leveraged to create realistic mutual spaces.

A benchmark of saliency models performance with a synthetic image dataset is provided. Model performance is evaluated through saliency metrics as well as the influence of model inspiration and consistency with human psychophysics. SID4VAM is composed of 230 synthetic images, with known salient regions. Images were generated with 15 distinct types of low-level features (e.g. orientation, brightness, color, size...) with a target-distractor pop-out type of synthetic patterns. We have used Free-Viewing and Visual Search task instructions and 7 feature contrasts for each feature category. Our study reveals that state-of-the-art Deep Learning saliency models do not perform well with synthetic pattern images, instead, models with Spectral/Fourier inspiration outperform others in saliency metrics and are more consistent with human psychophysical experimentation. This study proposes a new way to evaluate saliency models in the forthcoming literature, accounting for synthetic images with uniquely low-level feature contexts, distinct from previous eye tracking image datasets.

Artificially crafted images such as memes, seasonal greetings, etc are flooding the social media platforms today. These eventually start occupying a lot of internal memory of smartphones and it gets cumbersome for the user to go through hundreds of images and delete these synthetic images. To address this, we propose a novel method based on Convolutional Neural Networks (CNNs) for the on-device filtering of social media images by classifying these synthetic images and allowing the user to delete them in one go. The custom model uses depthwise separable convolution layers to achieve low inference time on smartphones. We have done an extensive evaluation of our model on various camera image datasets to cover most aspects of images captured by a camera. Various sorts of synthetic social media images have also been tested. The proposed solution achieves an accuracy of 98.25% on the Places-365 dataset and 95.81% on the Synthetic image dataset that we have prepared containing 30K instances.