-
Thesis for Master's Degree
13 Jun 2023
In this study, we aim to achieve lightweight models by adopting the Convolutional Vision Transformer (CvT) as the backbone and incorporating key techniques of Inpainting, namely update mask strategy and skip connections, to enhance the model’s performance. Experiments were conducted on the Tiny-ImageNet-200 and ImageNet- 1k datasets. The results of our research demonstrate the effectiveness of our approach in improving performance in Self-supervised Learning through model lightweighting and novel training strategies. This provides a promising direction for efficient model training under resource constraints. Our findings suggest the potential of a lightweight CvT model in the context of Inpainting tasks, offering valuable insights for researchers in the field of computer vision.
-
Medical Image Classification
12 Jun 2023
This study proposes a model that utilizes self-supervised learning with Convolutional Vision Transformer (CvT) to overcome the challenges of data collection and annotation in the medical field. This model has the potential for stable utilization in medical image classification tasks, even with small datasets, and is expected to significantly contribute to the improvement of performance in medical image analysis.