hovhannes22/ddpm
Unconditional DDPM for Armenian Letters
This repository implements an unconditional Denoising Diffusion Probabilistic Model (DDPM) trained on a custom dataset of 19,608 images of Armenian letters across 258 fonts. The project includes a UNet-based model, dataset preparation scripts, training code, and uses DPM-Solver for efficient inference. It generates grayscale 96x96 images of Armenian typography, sometimes creating novel glyphs.
A Jupyter Notebook implementing a ResNet-based deep learning model for Armenian handwritten character recognition, tr...
A Convolutional Neural Network (CNN) model for recognizing characters of the Armenian alphabet, achieving ~95% accura...
A deep learning project implementing convolutional neural networks (CNNs) to classify handwritten Armenian alphabet c...
A Jupyter Notebook project implementing a Convolutional Neural Network (CNN) to classify Armenian handwritten letters...
A project that trains a LeNet-5 convolutional neural network on Armenian script characters. It includes data preparat...
This repository implements an Armenian handwriting text recognition system using computer vision and machine learning...