FEWSAM Few-shot Segmentation tool based on Segment Anything

Use denoising diffusion model to segment the objects on the image step by step.

One way Few-shot Segmentation

A synthetic data generator using stable diffusion, it works with image inputs instead of text input

CLIP Feature Extractor and Linear Projection Loss for Deep Metric Learning

Clean and basic implementation of retinanet object detection.

A Self Supervised Learning Method using Linear Projection

A Self-supervised learning algorithm that uses summation and difference of embedding vectors that are generated from partial and full image.

Entangled state prediction with respect to time

Multi-Perspective Neural Networks(MPNN) applied on ImageNet dataset. MPNN is a unsupervised learning algorithm.

A CNN Model that creates filters bu unsupervised learning. The model tries to distinguish the output of each convolution operation and tries to create filters that generate most various output layers.

Experiments which are conducted in Section 4.1 in master thesis "Development of Deep Neural Networks that learns faster"

Experiments which are conducted in Section 5.4 in master thesis "Development of Deep Neural Networks that learns faster"

Feed-forward neural network model using w_oxsin(w_ixpi) instead of using w*x+b as basis function.