It is well known that the performance of any classification model is effective if the dataset used for the training process and the test process satisfy some specific requirements. In other words, the more the dataset size is large, balanced, and representative, the more one can trust the proposed model's effectiveness and, consequently, the obtained results. Unfortunately, large-size anonymous datasets are generally not publicly available in biomedical applications, especially those dealing with pathological human face images. This concern makes using deep-learning-based approaches challenging to deploy and difficult to reproduce or verify some published results. In this paper, we suggest an efficient method to generate a realistic anonymous synthetic dataset of human faces with the attributes of acne disorders corresponding to three levels of severity (i.e. Mild, Moderate and Severe). Therefore, a specific hierarchy StyleGAN-based algorithm trained at distinct levels is considered. To evaluate the performance of the proposed scheme, we consider a CNN-based classification system, trained using the generated synthetic acneic face images and tested using authentic face images. Consequently, we show that an accuracy of 97,6\% is achieved using InceptionResNetv2. As a result, this work allows the scientific community to employ the generated synthetic dataset for any data processing application without restrictions on legal or ethical concerns. Moreover, this approach can also be extended to other applications requiring the generation of synthetic medical images. We can make the code and the generated dataset accessible for the scientific community.
Missing data is a common concern in health datasets, and its impact on good decision-making processes is well documented. Our study's contribution is a methodology for tackling missing data problems using a combination of synthetic dataset generation, missing data imputation and deep learning methods to resolve missing data challenges. Specifically, we conducted a series of experiments with these objectives; $a)$ generating a realistic synthetic dataset, $b)$ simulating data missingness, $c)$ recovering the missing data, and $d)$ analyzing imputation performance. Our methodology used a gaussian mixture model whose parameters were learned from a cleaned subset of a real demographic and health dataset to generate the synthetic data. We simulated various missingness degrees ranging from $10 \%$, $20 \%$, $30 \%$, and $40\%$ under the missing completely at random scheme MCAR. We used an integrated performance analysis framework involving clustering, classification and direct imputation analysis. Our results show that models trained on synthetic and imputed datasets could make predictions with an accuracy of $83 \%$ and $80 \%$ on $a) $ an unseen real dataset and $b)$ an unseen reserved synthetic test dataset, respectively. Moreover, the models that used the DAE method for imputed yielded the lowest log loss an indication of good performance, even though the accuracy measures were slightly lower. In conclusion, our work demonstrates that using our methodology, one can reverse engineer a solution to resolve missingness on an unseen dataset with missingness. Moreover, though we used a health dataset, our methodology can be utilized in other contexts.
Recently, dataset-generation-based zero-shot learning has shown promising results by training a task-specific model with a dataset synthesized from large pre-trained language models (PLMs). The final task-specific model often achieves compatible or even better performance than PLMs under the zero-shot setting, with orders of magnitude fewer parameters. However, synthetic datasets have their drawbacks. They have long been suffering from low-quality issues (e.g., low informativeness and redundancy). This explains why the massive synthetic data does not lead to better performance -- a scenario we would expect in the human-labeled data. To improve the quality of dataset synthesis, we propose a progressive zero-shot dataset generation framework, ProGen, which leverages the feedback from the task-specific model to guide the generation of new training data via in-context examples. Extensive experiments on five text classification datasets demonstrate the effectiveness of the proposed approach. We also show ProGen achieves on-par or superior performance with only 1\% synthetic dataset size compared to baseline methods without in-context feedback.
State-of-the-art approaches in computer vision heavily rely on sufficiently large training datasets. For real-world applications, obtaining such a dataset is usually a tedious task. In this paper, we present a fully automated pipeline to generate a synthetic dataset for instance segmentation in four steps. In contrast to existing work, our pipeline covers every step from data acquisition to the final dataset. We first scrape images for the objects of interest from popular image search engines and since we rely only on text-based queries the resulting data comprises a wide variety of images. Hence, image selection is necessary as a second step. This approach of image scraping and selection relaxes the need for a real-world domain-specific dataset that must be either publicly available or created for this purpose. We employ an object-agnostic background removal model and compare three different methods for image selection: Object-agnostic pre-processing, manual image selection and CNN-based image selection. In the third step, we generate random arrangements of the object of interest and distractors on arbitrary backgrounds. Finally, the composition of the images is done by pasting the objects using four different blending methods. We present a case study for our dataset generation approach by considering parcel segmentation. For the evaluation we created a dataset of parcel photos that were annotated automatically. We find that (1) our dataset generation pipeline allows a successful transfer to real test images (Mask AP 86.2), (2) a very accurate image selection process - in contrast to human intuition - is not crucial and a broader category definition can help to bridge the domain gap, (3) the usage of blending methods is beneficial compared to simple copy-and-paste. We made our full code for scraping, image composition and training publicly available at https://a-nau.github.io/parcel2d.
The application of computer vision and machine learning methods in the field of additive manufacturing (AM) for semantic segmentation of the structural elements of 3-D printed products will improve real-time failure analysis systems and can potentially reduce the number of defects by enabling in situ corrections. This work demonstrates the possibilities of using physics-based rendering for labeled image dataset generation, as well as image-to-image translation capabilities to improve the accuracy of real image segmentation for AM systems. Multi-class semantic segmentation experiments were carried out based on the U-Net model and cycle generative adversarial network. The test results demonstrated the capacity of detecting such structural elements of 3-D printed parts as a top layer, infill, shell, and support. A basis for further segmentation system enhancement by utilizing image-to-image style transfer and domain adaptation technologies was also developed. The results indicate that using style transfer as a precursor to domain adaptation can significantly improve real 3-D printing image segmentation in situations where a model trained on synthetic data is the only tool available. The mean intersection over union (mIoU) scores for synthetic test datasets included 94.90% for the entire 3-D printed part, 73.33% for the top layer, 78.93% for the infill, 55.31% for the shell, and 69.45% for supports.
Despite tremendous progress in natural language processing using deep learning techniques in recent years, sign language production and comprehension has advanced very little. One critical barrier is the lack of largescale datasets available to the public due to the unbearable cost of labeled data generation. Efforts to provide public data for American Sign Language (ASL) comprehension have yielded two datasets, comprising more than thousand video clips. These datasets are large enough to enable a meaningful start to deep learning research on sign languages but are far too small to lead to any solution that can be practically deployed. So far, there is still no suitable dataset for ASL production. We proposed a system that can generate large scale ASL datasets for continuous ASL. It is suitable for general ASL processing and is particularly useful for ASL production. The continuous ASL dataset contains English labeled human articulations in condensed body pose data formats. To better serve the research community, we are releasing the first version of our ASL dataset, which contains 30k sentences, 416k words, a vocabulary of 18k words, in a total of 104 hours. This is the largest continuous sign language dataset published to date in terms of video duration. We also describe a system that can evolve and expand the dataset to incorporate better data processing techniques and more contents when available. It is our hope that the release of this ASL dataset and the sustainable dataset generation system to the public will propel better deep-learning research in ASL natural language processing.
Machine Learning (ML) has achieved enormous success in solving a variety of problems in computer vision, speech recognition, object detection, to name a few. The principal reason for this success is the availability of huge datasets for training deep neural networks (DNNs). However, datasets can not be publicly released if they contain sensitive information such as medical or financial records. In such cases, data privacy becomes a major concern. Encryption methods offer a possible solution to this issue, however their deployment on ML applications is non-trivial, as they seriously impact the classification accuracy and result in substantial computational overhead.Alternatively, obfuscation techniques can be used, but maintaining a good balance between visual privacy and accuracy is challenging. In this work, we propose a method to generate secure synthetic datasets from the original private datasets. In our method, given a network with Batch Normalization (BN) layers pre-trained on the original dataset, we first record the layer-wise BN statistics. Next, using the BN statistics and the pre-trained model, we generate the synthetic dataset by optimizing random noises such that the synthetic data match the layer-wise statistical distribution of the original model. We evaluate our method on image classification dataset (CIFAR10) and show that our synthetic data can be used for training networks from scratch, producing reasonable classification performance.
Recently there has been increasing interest in developing and deploying deep graph learning algorithms for many tasks, such as fraud detection and recommender systems. Albeit, there is a limited number of publicly available graph-structured datasets, most of which are tiny compared to production-sized applications or are limited in their application domain. This work tackles this shortcoming by proposing a scalable synthetic graph generation tool to scale the datasets to production-size graphs with trillions of edges and billions of nodes. The tool learns a series of parametric models from proprietary datasets that can be released to researchers to study various graph methods on the synthetic data increasing prototype development and novel applications. We demonstrate the generalizability of the framework across a series of datasets, mimicking structural and feature distributions as well as the ability to scale them across varying sizes demonstrating their usefulness for benchmarking and model development. Code can be found on https://github.com/NVIDIA/DeepLearningExamples/tree/master/Tools/DGLPyTorch/SyntheticGraphGeneration.
Noise in seismic data arises from numerous sources and is continually evolving. The use of supervised deep learning procedures for denoising of seismic datasets often results in poor performance: this is due to the lack of noise-free field data to act as training targets and the large difference in characteristics between synthetic and field datasets. Self-supervised, blind-spot networks typically overcome these limitation by training directly on the raw, noisy data. However, such networks often rely on a random noise assumption, and their denoising capabilities quickly decrease in the presence of even minimally-correlated noise. Extending from blind-spots to blind-masks can efficiently suppress coherent noise along a specific direction, but it cannot adapt to the ever-changing properties of noise. To preempt the network's ability to predict the signal and reduce its opportunity to learn the noise properties, we propose an initial, supervised training of the network on a frugally-generated synthetic dataset prior to fine-tuning in a self-supervised manner on the field dataset of interest. Considering the change in peak signal-to-noise ratio, as well as the volume of noise reduced and signal leakage observed, we illustrate the clear benefit in initialising the self-supervised network with the weights from a supervised base-training. This is further supported by a test on a field dataset where the fine-tuned network strikes the best balance between signal preservation and noise reduction. Finally, the use of the unrealistic, frugally-generated synthetic dataset for the supervised base-training includes a number of benefits: minimal prior geological knowledge is required, substantially reduced computational cost for the dataset generation, and a reduced requirement of re-training the network should recording conditions change, to name a few.
The availability of real data from areas with high privacy requirements, such as the medical intervention space, is low and the acquisition legally complex. Therefore, this work presents a way to create a synthetic dataset for the medical context, using medical clothing as an example. The goal is to close the reality gap between the synthetic and real data. For this purpose, methods of 3D-scanned clothing and designed clothing are compared in a Domain-Randomization and Structured-Domain-Randomization scenario using an Unreal-Engine plugin or Unity. Additionally a Mixed-Reality dataset in front of a greenscreen and a target domain dataset were used. Our experiments show, that Structured-Domain-Randomization of designed clothing together with Mixed-Reality data provide a baseline achieving 72.0% mAP on a test dataset of the clinical target domain. When additionally using 15% of available target domain train data, the gap towards 100% (660 images) target domain train data could be nearly closed 80.05% mAP (81.95% mAP). Finally we show that when additionally using 100% target domain train data the accuracy could be increased to 83.35% mAP.
Conversational question--answer generation is a task that automatically generates a large-scale conversational question answering dataset based on input passages. In this paper, we introduce a novel framework that extracts question-worthy phrases from a passage and then generates corresponding questions considering previous conversations. In particular, our framework revises the extracted answers after generating questions so that answers exactly match paired questions. Experimental results show that our simple answer revision approach leads to significant improvement in the quality of synthetic data. Moreover, we prove that our framework can be effectively utilized for domain adaptation of conversational question answering.
Datasets are essential to apply AI algorithms to Cyber Physical System (CPS) Security. Due to scarcity of real CPS datasets, researchers elected to generate their own datasets using either real or virtualized testbeds. However, unlike other AI domains, a CPS is a complex system with many interfaces that determine its behavior. A dataset that comprises merely a collection of sensor measurements and network traffic may not be sufficient to develop resilient AI defensive or offensive agents. In this paper, we study the \emph{elements} of CPS security datasets required to capture the system behavior and interactions, and propose a dataset architecture that has the potential to enhance the performance of AI algorithms in securing cyber physical systems. The framework includes dataset elements, attack representation, and required dataset features. We compare existing datasets to the proposed architecture to identify the current limitations and discuss the future of CPS dataset generation using testbeds.
Environmental perception is an important aspect within the field of autonomous vehicles that provides crucial information about the driving domain, including but not limited to identifying clear driving areas and surrounding obstacles. Semantic segmentation is a widely used perception method for self-driving cars that associates each pixel of an image with a predefined class. In this context, several segmentation models are evaluated regarding accuracy and efficiency. Experimental results on the generated dataset confirm that the segmentation model FasterSeg is fast enough to be used in realtime on lowpower computational (embedded) devices in self-driving cars. A simple method is also introduced to generate synthetic training data for the model. Moreover, the accuracy of the first-person perspective and the bird's eye view perspective are compared. For a $320 \times 256$ input in the first-person perspective, FasterSeg achieves $65.44\,\%$ mean Intersection over Union (mIoU), and for a $320 \times 256$ input from the bird's eye view perspective, FasterSeg achieves $64.08\,\%$ mIoU. Both perspectives achieve a frame rate of $247.11$ Frames per Second (FPS) on the NVIDIA Jetson AGX Xavier. Lastly, the frame rate and the accuracy with respect to the arithmetic 16-bit Floating Point (FP16) and 32-bit Floating Point (FP32) of both perspectives are measured and compared on the target hardware.
Obtaining the ground truth labels from a video is challenging since the manual annotation of pixel-wise flow labels is prohibitively expensive and laborious. Besides, existing approaches try to adapt the trained model on synthetic datasets to authentic videos, which inevitably suffers from domain discrepancy and hinders the performance for real-world applications. To solve these problems, we propose RealFlow, an Expectation-Maximization based framework that can create large-scale optical flow datasets directly from any unlabeled realistic videos. Specifically, we first estimate optical flow between a pair of video frames, and then synthesize a new image from this pair based on the predicted flow. Thus the new image pairs and their corresponding flows can be regarded as a new training set. Besides, we design a Realistic Image Pair Rendering (RIPR) module that adopts softmax splatting and bi-directional hole filling techniques to alleviate the artifacts of the image synthesis. In the E-step, RIPR renders new images to create a large quantity of training data. In the M-step, we utilize the generated training data to train an optical flow network, which can be used to estimate optical flows in the next E-step. During the iterative learning steps, the capability of the flow network is gradually improved, so is the accuracy of the flow, as well as the quality of the synthesized dataset. Experimental results show that RealFlow outperforms previous dataset generation methods by a considerably large margin. Moreover, based on the generated dataset, our approach achieves state-of-the-art performance on two standard benchmarks compared with both supervised and unsupervised optical flow methods. Our code and dataset are available at https://github.com/megvii-research/RealFlow
Existing adversarial example research focuses on digitally inserted perturbations on top of existing natural image datasets. This construction of adversarial examples is not realistic because it may be difficult, or even impossible, for an attacker to deploy such an attack in the real-world due to sensing and environmental effects. To better understand adversarial examples against cyber-physical systems, we propose approximating the real-world through simulation. In this paper we describe our synthetic dataset generation tool that enables scalable collection of such a synthetic dataset with realistic adversarial examples. We use the CARLA simulator to collect such a dataset and demonstrate simulated attacks that undergo the same environmental transforms and processing as real-world images. Our tools have been used to collect datasets to help evaluate the efficacy of adversarial examples, and can be found at https://github.com/carla-simulator/carla/pull/4992.
In the past decade, object detection tasks are defined mostly by large public datasets. However, building object detection datasets is not scalable due to inefficient image collecting and labeling. Furthermore, most labels are still in the form of bounding boxes, which provide much less information than the real human visual system. In this paper, we present a method to synthesize object-in-scene images, which can preserve the objects' detailed features without bringing irrelevant information. In brief, given a set of images containing a target object, our algorithm first trains a model to find an approximate center of the object as an anchor, then makes an outline regression to estimate its boundary, and finally blends the object into a new scene. Our result shows that in the synthesized image, the boundaries of objects blend very well with the background. Experiments also show that SOTA segmentation models work well with our synthesized data.
Reconstructing 3D objects is an important computer vision task that has wide application in AR/VR. Deep learning algorithm developed for this task usually relies on an unrealistic synthetic dataset, such as ShapeNet and Things3D. On the other hand, existing real-captured object-centric datasets usually do not have enough annotation to enable supervised training or reliable evaluation. In this technical report, we present a photo-realistic object-centric dataset HM3D-ABO. It is constructed by composing realistic indoor scene and realistic object. For each configuration, we provide multi-view RGB observations, a water-tight mesh model for the object, ground truth depth map and object mask. The proposed dataset could also be useful for tasks such as camera pose estimation and novel-view synthesis. The dataset generation code is released at https://github.com/zhenpeiyang/HM3D-ABO.
To assist in the development of machine learning methods for automated classification of spectroscopic data, we have generated a universal synthetic dataset that can be used for model validation. This dataset contains artificial spectra designed to represent experimental measurements from techniques including X-ray diffraction, nuclear magnetic resonance, and Raman spectroscopy. The dataset generation process features customizable parameters, such as scan length and peak count, which can be adjusted to fit the problem at hand. As an initial benchmark, we simulated a dataset containing 35,000 spectra based on 500 unique classes. To automate the classification of this data, eight different machine learning architectures were evaluated. From the results, we shed light on which factors are most critical to achieve optimal performance for the classification task. The scripts used to generate synthetic spectra, as well as our benchmark dataset and evaluation routines, are made publicly available to aid in the development of improved machine learning models for spectroscopic analysis.
In general, to draw robust conclusions from a dataset, all the analyzed population must be represented on said dataset. Having a dataset that does not fulfill this condition normally leads to selection bias. Additionally, graphs have been used to model a wide variety of problems. Although synthetic graphs can be used to augment available real graph datasets to overcome selection bias, the generation of unbiased synthetic datasets is complex with current tools. In this work, we propose a method to find a synthetic graph dataset that has an even representation of graphs with different metrics. The resulting dataset can then be used, among others, for benchmarking graph processing techniques as the accuracy of different Graph Neural Network (GNN) models or the speedups obtained by different graph processing acceleration frameworks.
Recent methods for visual question answering rely on large-scale annotated datasets. Manual annotation of questions and answers for videos, however, is tedious, expensive and prevents scalability. In this work, we propose to avoid manual annotation and generate a large-scale training dataset for video question answering making use of automatic cross-modal supervision. We leverage a question generation transformer trained on text data and use it to generate question-answer pairs from transcribed video narrations. Given narrated videos, we then automatically generate the HowToVQA69M dataset with 69M video-question-answer triplets. To handle the open vocabulary of diverse answers in this dataset, we propose a training procedure based on a contrastive loss between a video-question multi-modal transformer and an answer transformer. We introduce the zero-shot VideoQA task and the VideoQA feature probe evaluation setting and show excellent results, in particular for rare answers. Furthermore, our method achieves competitive results on MSRVTT-QA, ActivityNet-QA, MSVD-QA and How2QA datasets. We also show that our VideoQA dataset generation approach generalizes to another source of web video and text data. We use our method to generate the WebVidVQA3M dataset from the WebVid dataset, i.e., videos with alt-text annotations, and show its benefits for training VideoQA models. Finally, for a detailed evaluation we introduce iVQA, a new VideoQA dataset with reduced language bias and high-quality manual annotations. Code, datasets and trained models are available at https://antoyang.github.io/just-ask.html