Ivana Marin - Academia.edu (original) (raw)
Uploads
Papers by Ivana Marin
2020 International Conference on Software, Telecommunications and Computer Networks (SoftCOM)
Meaningful and objective evaluation metric for fair model comparison is crucial for further scien... more Meaningful and objective evaluation metric for fair model comparison is crucial for further scientific progress in the field of deep generative modeling. Despite the significant progress and impressive results obtained by Generative Adversarial Networks in recent years, the problem of their objective evaluation remains open. In this paper, we give an overview of qualitative and quantitative evaluation measures most frequently used to assess the quality of generated images and learned representations of an adversarial network together with the empirical comparison of their performance on the problem of human face image synthesis. It is shown that evaluation scores of the two most widely accepted quantitative metrics, Inception Score (IS) and Fréchet Inception Distance (FID), do not correlate. The IS is not an appropriate evaluation metric for a given problem, but FID shows good performance that correlates well with a visual inspection of generated samples. The qualitative evaluation can be used to complement results obtained with quantitative evaluation - to gain further insight into the learned data representation and detect possible overfitting.
Journal of communications software and systems, 2021
In recent years Generative Adversarial Networks (GANs) have achieved remarkable results in the ta... more In recent years Generative Adversarial Networks (GANs) have achieved remarkable results in the task of realistic image synthesis. Despite their continued success and advances, there still lacks a thorough understanding of how precisely GANs map random latent vectors to realistic-looking images and how the priors set on the latent space affect the learned mapping. In this work, we analyze the effect of the chosen latent dimension on the final quality of synthesized images of human faces and learned data representations. We show that GANs can generate images plausibly even with latent dimensions significantly smaller than the standard dimensions like 100 or 512. Although one might expect that larger latent dimensions encourage the generation of more diverse and enhanced quality images, we show that an increase of latent dimension after some point does not lead to visible improvements in perceptual image quality nor in quantitative estimates of its generalization abilities.
The global community has recognized an increasing amount of pollutants entering oceans and other ... more The global community has recognized an increasing amount of pollutants entering oceans and other water bodies as a severe environmental, economic, and social issue. In addition to prevention, one of the key measures in addressing marine pollution is the cleanup of debris already present in marine environments. Deployment of machine learning (ML) and deep learning (DL) techniques can automate marine waste removal, making the cleanup process more efficient. This study examines the performance of six well-known deep convolutional neural networks (CNNs), namely VGG19, InceptionV3, ResNet50, Inception-ResNetV2, DenseNet121, and MobileNetV2, utilized as feature extractors according to three different extraction schemes for the identification and classification of underwater marine debris. We compare the performance of a neural network (NN) classifier trained on top of deep CNN feature extractors when the feature extractor is (1) fixed; (2) fine-tuned on the given task; (3) fixed during th...
Applied Sciences
The main goal of any classification or regression task is to obtain a model that will generalize ... more The main goal of any classification or regression task is to obtain a model that will generalize well on new, previously unseen data. Due to the recent rise of deep learning and many state-of-the-art results obtained with deep models, deep learning architectures have become one of the most used model architectures nowadays. To generalize well, a deep model needs to learn the training data well without overfitting. The latter implies a correlation of deep model optimization and regularization with generalization performance. In this work, we explore the effect of the used optimization algorithm and regularization techniques on the final generalization performance of the model with convolutional neural network (CNN) architecture widely used in the field of computer vision. We give a detailed overview of optimization and regularization techniques with a comparative analysis of their performance with three CNNs on the CIFAR-10 and Fashion-MNIST image datasets.
2020 International Conference on Software, Telecommunications and Computer Networks (SoftCOM)
Meaningful and objective evaluation metric for fair model comparison is crucial for further scien... more Meaningful and objective evaluation metric for fair model comparison is crucial for further scientific progress in the field of deep generative modeling. Despite the significant progress and impressive results obtained by Generative Adversarial Networks in recent years, the problem of their objective evaluation remains open. In this paper, we give an overview of qualitative and quantitative evaluation measures most frequently used to assess the quality of generated images and learned representations of an adversarial network together with the empirical comparison of their performance on the problem of human face image synthesis. It is shown that evaluation scores of the two most widely accepted quantitative metrics, Inception Score (IS) and Fréchet Inception Distance (FID), do not correlate. The IS is not an appropriate evaluation metric for a given problem, but FID shows good performance that correlates well with a visual inspection of generated samples. The qualitative evaluation can be used to complement results obtained with quantitative evaluation - to gain further insight into the learned data representation and detect possible overfitting.
Journal of communications software and systems, 2021
In recent years Generative Adversarial Networks (GANs) have achieved remarkable results in the ta... more In recent years Generative Adversarial Networks (GANs) have achieved remarkable results in the task of realistic image synthesis. Despite their continued success and advances, there still lacks a thorough understanding of how precisely GANs map random latent vectors to realistic-looking images and how the priors set on the latent space affect the learned mapping. In this work, we analyze the effect of the chosen latent dimension on the final quality of synthesized images of human faces and learned data representations. We show that GANs can generate images plausibly even with latent dimensions significantly smaller than the standard dimensions like 100 or 512. Although one might expect that larger latent dimensions encourage the generation of more diverse and enhanced quality images, we show that an increase of latent dimension after some point does not lead to visible improvements in perceptual image quality nor in quantitative estimates of its generalization abilities.
The global community has recognized an increasing amount of pollutants entering oceans and other ... more The global community has recognized an increasing amount of pollutants entering oceans and other water bodies as a severe environmental, economic, and social issue. In addition to prevention, one of the key measures in addressing marine pollution is the cleanup of debris already present in marine environments. Deployment of machine learning (ML) and deep learning (DL) techniques can automate marine waste removal, making the cleanup process more efficient. This study examines the performance of six well-known deep convolutional neural networks (CNNs), namely VGG19, InceptionV3, ResNet50, Inception-ResNetV2, DenseNet121, and MobileNetV2, utilized as feature extractors according to three different extraction schemes for the identification and classification of underwater marine debris. We compare the performance of a neural network (NN) classifier trained on top of deep CNN feature extractors when the feature extractor is (1) fixed; (2) fine-tuned on the given task; (3) fixed during th...
Applied Sciences
The main goal of any classification or regression task is to obtain a model that will generalize ... more The main goal of any classification or regression task is to obtain a model that will generalize well on new, previously unseen data. Due to the recent rise of deep learning and many state-of-the-art results obtained with deep models, deep learning architectures have become one of the most used model architectures nowadays. To generalize well, a deep model needs to learn the training data well without overfitting. The latter implies a correlation of deep model optimization and regularization with generalization performance. In this work, we explore the effect of the used optimization algorithm and regularization techniques on the final generalization performance of the model with convolutional neural network (CNN) architecture widely used in the field of computer vision. We give a detailed overview of optimization and regularization techniques with a comparative analysis of their performance with three CNNs on the CIFAR-10 and Fashion-MNIST image datasets.