Global ETD Search

Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.

Results

Showing 1 to 20 of 28 for “"Self-supervised pretraining"”.

  1. Self-supervised Pretraining of Vision Transformers for Earth Observation

    … foundation models for Earth Observation by pretraining vision transformers (ViTs) using self-supervised learning. Using masked autoencoding, I develop one of the first ViTs pretrained on RS data—called SatViT. Next, I develop an improved foundation model—called SatViT-V2—and evaluate it on …

    carleton Repository record for Self-supervised Pretraining of Vision Transformers for Earth Observation (opens in a new tab)

  2. Towards Robust Deep Learning for Medical Imaging with Limited and Noisy Labeled Data

    … surgical intervention, and treatment planning. Supervised deep learning methods typically require large amounts of high-quality labeled data for training. However, acquiring high-quality labeled data in the medical domain is challenging due to factors such as the high cost of expert annotation …

    rit Repository record for Towards Robust Deep Learning for Medical Imaging with Limited and Noisy Labeled Data (opens in a new tab)

  3. Self-supervised přístupy pro efektivní analýzu 3D tvarů

    … těchto dat. Při adaptaci těchto modelů pro self-supervised pretraining pomocí autoencoderů je identifikován zásadní nedostatek: stávající pooling a zejména unpooling metody nejsou dostatečné pro kvalitní rekonstrukci. Tato príce prezentuje novou statickou předpočítanou pooling a unpooling …

    brno-tech Repository record for Self-supervised přístupy pro efektivní analýzu 3D tvarů (opens in a new tab)

  4. Protecting wild Atlantic salmon : using tiny data to specialize visual transformer foundation models for image classification

    … with image classification. The performance of supervised adaptation of a ViT foundation model to a tiny, task-specific dataset is evaluated and compared between model size and parameters. The effects of Low Rank Adaptation of a ViT foundation model when adapting to a tiny, task-specific dataset …

    reykjavik Repository record for Protecting wild Atlantic salmon : using tiny data to specialize visual transformer foundation models for image classification (opens in a new tab)

  5. VISUAL RELATION DRIVEN VIDEO QUESTION ANSWERING

    … as learning from vast amounts of Web data via self-supervised pretraining. With these efforts, we hope this thesis can provide a foundation for future studies in relation-aware videoQA.

    nus Repository record for VISUAL RELATION DRIVEN VIDEO QUESTION ANSWERING (opens in a new tab) Download file for VISUAL RELATION DRIVEN VIDEO QUESTION ANSWERING (opens in a new tab)

  6. Knowledge-Informed Weakly-Supervised Deep Learning Models for Cancer Applications

    … heterogeneity in cancer imaging. They combine self-supervised pretraining, knowledge-informed loss functions, hierarchical and contextualized architectures, and label smoothing techniques that distill clinical and biological priors. These approaches enable dense spatial prediction of gene …

    gatech Repository record for Knowledge-Informed Weakly-Supervised Deep Learning Models for Cancer Applications (opens in a new tab)

  7. Scene Change Detection

    … learning from large ImageNet datasets. This is a supervised technique and to overcome the challenges mentioned above, we introduce a self-supervised pretraining method with unlabeled datasets based on a existing D-SSCD approach that learns temporal-consistent representations of a pair of images. …

    chalmers Repository record for Scene Change Detection (opens in a new tab)

  8. Learning deep neural networks: necessity and scope of prior knowledge, raw data, and labels

    … supervisory signals are scarce. The field of self-supervised learning has therefore become an increasingly popular framework for learning without labels and in this thesis I provide several works that shed light on the workings of self-supervised learning, and that contributed to, and shaped …

    oxford Repository record for Learning deep neural networks: necessity and scope of prior knowledge, raw data, and labels (opens in a new tab)

  9. REDUCING THE BURDEN OF CANCER WITH ARTIFICIAL INTELLIGENCE: IMAGE-BASED MODELS FOR BREAST CANCER DETECTION, ADVANCED STAGE RISK, AND CROSS-MODALITY VISUALIZATION

    … of advanced-stage breast cancer; 3) Implement pretraining methods to improve modeling with limited labeled medical images for body composition. Novel contributions to the field resulting from each aim are as follows: 1) Malignant lesions were confirmed to have unique compositional signatures …

    hawaii Repository record for REDUCING THE BURDEN OF CANCER WITH ARTIFICIAL INTELLIGENCE: IMAGE-BASED MODELS FOR BREAST CANCER DETECTION, ADVANCED STAGE RISK, AND CROSS-MODALITY VISUALIZATION (opens in a new tab)

  10. Inferring Spatiotemporal Structure through Self-Supervised Learning: From Predictive Coding to Neuro Foundation Models

    … In this thesis, I investigate prediction-based self-supervised learning as a principle for learning such structure in both biological and artificial systems. I first study sparse coding variational autoencoders as a bridge between classical predictive coding models and modern self-supervised …

    washington Repository record for Inferring Spatiotemporal Structure through Self-Supervised Learning: From Predictive Coding to Neuro Foundation Models (opens in a new tab)

  11. Advanced Deep Learning Algorithms for Medical Imaging Applications: Tumor and Organ at Risk Segmentation

    … performance in ambiguous regions. Additionally, self-supervised pretraining is employed to improve feature representation and generalization, particularly in settings with limited labeled data. The second component examines how tumor-specific characteristics relate to segmentation performance. …

    ucf Repository record for Advanced Deep Learning Algorithms for Medical Imaging Applications: Tumor and Organ at Risk Segmentation (opens in a new tab)

  12. Improving the Generalisation of Neural Networks with Unlabelled Data

    … However, in order to achieve good performance on supervised learning tasks, neural networks have to be trained on large amounts of labelled training data. This often requires human experts to painstakingly label every single example in the data. The labelling process can be costly and time …

    auckland-ms Repository record for Improving the Generalisation of Neural Networks with Unlabelled Data (opens in a new tab)

  13. Toward robust and fair ultrasound classification with clinical context

    … and enhance representation stability through self-supervised pretraining and prototype-based architectures. These methods improve diagnostic accuracy and robustness in prostate ultrasound cancer detection. Second, we address the absence of patient-specific contextual reasoning in current …

    ubc Repository record for Toward robust and fair ultrasound classification with clinical context (opens in a new tab) Download PDF for Toward robust and fair ultrasound classification with clinical context (opens in a new tab)

  14. From edges to pages: boundary-aware binarization and two-stage reconstruction of historical documents

    … future directions include multispectral fusion, self-supervised pretraining, and stronger contentpreservation priors.

    ets-quebec Download PDF for From edges to pages: boundary-aware binarization and two-stage reconstruction of historical documents (opens in a new tab)

  15. Deep Learning Pipelines for Brain-Wide Mapping of Neuronal Activity and Connectivity in Tera-Voxel Microscopy

    … voxel- and patch-level attentions, combined with self-supervised pretraining, and optimized augmentation and training strategies. Across multiple datasets, MAPL3 outperformed SOTA methods, demonstrating robustness to imaging artifacts and protocol variability, and producing anatomically accurate …

    toronto-retro Repository record for Deep Learning Pipelines for Brain-Wide Mapping of Neuronal Activity and Connectivity in Tera-Voxel Microscopy (opens in a new tab)

  16. Medical Image Segmentation Through Deep Learning on Companion Animal Thoracic Radiographs and MR Scans of the Human Brain

    … of new lesion segmentation models, we propose a self-supervised pretraining scheme based on image inpainting and synthetic lesion augmentation to increase the variability of samples our model is exposed to during training. Experiments conducted on the MSSEG-2 dataset demonstrate that our …

    calgary Repository record for Medical Image Segmentation Through Deep Learning on Companion Animal Thoracic Radiographs and MR Scans of the Human Brain (opens in a new tab)

  17. Facial Motion Augmented Identity Verification with Deep Neural Networks

    … on complex facial motions and proposed a new self-supervised pretraining approach to boost the encoder's performance. At last, we evaluated two other potential upstream tasks that could help improve the efficiency of facial motion encoding. Through these efforts, we have built a solid …

    byu Repository record for Facial Motion Augmented Identity Verification with Deep Neural Networks (opens in a new tab) Download PDF for Facial Motion Augmented Identity Verification with Deep Neural Networks (opens in a new tab)

  18. Robust and generalizable deep geometric representation learning for 3D point clouds

    … to acquire. This motivates the development of self-supervised learning methods that can pretrain models on unlabeled data and learn transferable geometric representations for downstream tasks with limited annotations. Second, models trained under fixed conditions often fail to generalize when …

    ets-quebec Download PDF for Robust and generalizable deep geometric representation learning for 3D point clouds (opens in a new tab)

  19. Data-efficient Semantic Segmentation with Deep Learning

    … we demonstrate a novel selective masking based self-supervised learning method suitable for CNNs. We improve on the inefficient random masking process that is widely used in masked image modelling by leveraging model knowledge to perform selective masking. We show on two weed segmentation …

    sask Repository record for Data-efficient Semantic Segmentation with Deep Learning (opens in a new tab)

Page 1 of 2