Active Image Synthesis for Efficient Labeling

Chen, Jialei; Xie, Yujia; Wang, Kan; Zhang, Chuck; Vannan, Mani A.; Wang, Ben; Qian, Zhen

doi:10.1109/tpami.2020.2993221

Cited by 11 publications

(4 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, we are also interested in monitoring cell manufacturing based on cell morphology. In this case, physics-informed deep learning frameworks in the literature (Raissi et al, 2017;Chen et al, 2020) appear to be suitable for recovering critical quality attributes in cell manufacturing.…”

Section: Discussionmentioning

confidence: 99%

A calibration-free method for biosensing in cell manufacturing

Chen,

Liu,

Wang

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Chimeric antigen receptor T cell therapy has demonstrated innovative therapeutic effectiveness in fighting cancers; however, it is extremely expensive due to the intrinsic patient-to-patient variability in cell manufacturing. We propose in this work a novel calibration-free statistical framework to effectively recover critical quality attributes under the patient-to-patient variability. Specifically, we model this variability via a patient-specific calibration parameter, and use readings from multiple biosensors to construct a patient-invariance statistic, thereby alleviating the effect of the calibration parameter. A carefully formulated optimization problem and an algorithmic framework are presented to find the best patient-invariance statistic and the model parameters. Using the patient-invariance statistic, we can recover the critical quality attribute of interest, free from the calibration parameter. We demonstrate improvements of the proposed calibration-free method in different simulation experiments. In the cell manufacturing case study, our method not only effectively recovers viable cell concentration for monitoring, but also reveals insights for the cell manufacturing process.

show abstract

Section: Discussionmentioning

confidence: 99%

A calibration-free method for biosensing in cell manufacturing

Chen,

Liu,

Wang

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The authors of [44] synthesize images of skin lesions to supplement the data using style-based GAN, eventually improving classification accuracy. To enhance the performance of limited data learning tasks, the research authors [11] present an AISEL, or active image synthesis approach for adequate labelling. The authors highlight a key element of AISEL known as the Generative Invertible Network (GIN), which generates physically accurate virtual images by extracting interpretable information from training images.…”

Section: Data Augmentation With Gansmentioning

confidence: 99%

Generative adversarial network based synthetic data training model for lightweight convolutional neural networks

Rather

Kumar

2023

Multimed Tools Appl

View full text Add to dashboard Cite

Inadequate training data is a significant challenge for deep learning techniques, particularly in applications where data is difficult to get, and publicly available datasets are uncommon owing to ethical and privacy concerns. Various approaches, such as data augmentation and transfer learning, are employed to address this problem, which help to some extent in removing this limitation. However, after a certain amount of data augmentation, the quality of the generated data stalls, and transfer learning suffers from the issue of negative transfer. This paper proposes a novel generative adversarial network-based synthetic data training (GAN-ST) model to generate synthetic data for training a lightweight convolutional neural network (CNN). An enhanced generator is proposed to quickly saturate and cover the colour space of the training distribution. The GAN-ST model is based on Deep Convolutional Generative Adversarial Network(s) (DCGAN) and Conditional Generative Adversarial Network(s) (CGAN) models, which consist of an enhanced generator. The study evaluates the accuracy of a CNN model on the MNIST and CIFAR 10 datasets using both original and synthetic data. The results revealed an impressive classifier accuracy on the MNIST dataset, achieving an accuracy of 99.38% on GAN-ST-generated synthetic training data, which is only 0.05% lower than the performance on original data-based training. The classifier performance on the CIFAR dataset is also remarkable, achieving an accuracy of 90.23%. The performance of CNN trained using GAN-ST-based synthetic data is notable, with the most considerable improvement of 0.66% and 7.06%, over a single GAN-based synthetic data training for the MNIST and CIFAR datasets, respectively. By training two GANs independently, the GAN-ST model covers different parts of the original data distribution, resulting in a more diverse and realistic training data set for the classifier. This diverse set of synthetic data, when used to train a CNN, shows better generalization to new data, leading to improved classification accuracy.

show abstract

“…The boundary condition of temperature is difficult to specify since the wafer is placed on a rotating platform with unknown temperature and complex heat flux. Another example is to understand the flood flow in healthcare applications [31,5]. While the governing PDEs are known, i.e., the Naiver-Stocks equation, the boundary conditions are extremely difficult to obtain, considering the patient-specific blood vessel geometry and the interaction between blood flow and soft biological tissues.…”

Section: Difference To Numerically Solving Pdesmentioning

confidence: 99%

APIK: Active Physics-Informed Kriging Model with Partial Differential Equations

Chen¹,

Zhang²,

Wu³

2020

Preprint

Self Cite

View full text Add to dashboard Cite

Kriging (or Gaussian process regression) is a popular machine learning method for its flexibility and closed-form prediction expressions. However, one of the key challenges in applying kriging to engineering systems is that the available measurement data is scarce due to the measurement limitations and high sensing costs. On the other hand, physical knowledge of the engineering system is often available and represented in the form of partial differential equations (PDEs). We present in this work a PDE Informed Kriging model (PIK), which introduces PDE information via a set of PDE points and conducts posterior prediction similar to the standard kriging method. The proposed PIK model can incorporate physical knowledge from both linear and nonlinear PDEs. To further improve learning performance, we propose an Active PIK framework (APIK) that designs PDE points to leverage the PDE information based on the PIK model and measurement data. The selected PDE points not only explore the whole input space but also exploit the locations where the PDE information is critical in reducing predictive uncertainty. Finally, an expectation-maximization algorithm is developed for parameter estimation. We demonstrate the effectiveness of APIK in two synthetic examples, a shock wave case study, and a laser heating case study.

show abstract

Active Image Synthesis for Efficient Labeling

Cited by 11 publications

References 45 publications

A calibration-free method for biosensing in cell manufacturing

A calibration-free method for biosensing in cell manufacturing

Generative adversarial network based synthetic data training model for lightweight convolutional neural networks

APIK: Active Physics-Informed Kriging Model with Partial Differential Equations

Contact Info

Product

Resources

About