You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A web-based tool that recognizes handwritten digits and alphabets using a Convolutional Neural Network (CNN) trained on the MNIST and EMNIST datasets. The system is built with Flask and TensorFlow, offering an intuitive interface for character recognition.
This repository contains a custom Arabic digits (0-9) dataset contributed by multiple individuals and a neural network model designed to accurately recognize these digits.
A "Hello World" ML neural network project features a FastAPI docker image for digit predictions and a React frontend where users can draw digits to see instant predictions
The MNIST dataset was used to train a neural network having a single linear layer with SoftMax employed in the criterion function (Cross Entropy Loss) to classify handwritten digits in classes 0 to 9. The model yielded a 92% accuracy on the MNIST test dataset in 10 training epochs.
The main goal of the project was to develop a model that accuratelyclassifies handwritten digits. I trained the model using the MNISTdataset and the CNN algorithm.. Later on, I integrated this model with image processingtechniques to create a project that recognizes digits in real-time
Este repositorio complementa mi Trabajo de Fin de Grado, que ha consistido en entrenar varios modelos de aprendizaje automático para que sean capaces de clasificar imágenes de símbolos matemáticos escritos a mano. Se han utilizado imágenes de la base de datos HASY, que cuenta con símbolos de 369 clases distintas.