Developed by Andreas Müller as part of a master thesis, WebNet Builder is an innovative No-Code platform for constructing neural networks directly within a web browser. Leveraging TensorFlow.js, it offers a user-friendly interface for comprehensive neural network development, with key features including:
- Interactive Neural Network Architecture Design: Utilize a drag-and-drop interface for building custom neural network models.
- Flexible Project Management: Seamlessly manage your projects with features like ZIP file import and export.
- Data Handling Capabilities: Import CSV datasets and perform necessary data preprocessing.
- Browser-Based Training: Train neural networks in-browser using various computational backends such as CPU, Web Assembly, WebGL and WebGPU.
- Training Progress Visualization: Monitor and visualize the training process of neural networks.
- Model Evaluation: Assess and compare the performance of different trained neural network models.
WebNet Builder democratizes neural network development, making it accessible and efficient for users without coding expertise.