kitsakisG/automl-forge

📸 AutoML Forge - Computer Vision Demo

State-of-the-art image classification with Vision Transformers and CNNs.

🎯 Try It Out

1. 1.Select a model from the dropdown (MobileNetV3, ResNet18, EfficientNet-B0, or ViT-Base)
2. 2.Upload an image (JPG, PNG)
3. 3.Click Predict to get top-5 predictions with confidence scores

🤖 Models Available

• —MobileNetV3: Fast, mobile-optimized (5.4M parameters)
• —ResNet18: Lightweight CNN (11.7M parameters)
• —EfficientNet-B0: Balanced speed/accuracy (5.3M parameters)
• —ViT-Base: State-of-the-art Vision Transformer (86M parameters)

All models are pre-trained on ImageNet (1.2M images, 1000 classes).

🚀 Full Platform

This is a simplified demo. The full AutoML Forge platform includes:

• —✅ Custom Training: Train on your own datasets
• —✅ AutoML: Automatic model selection and comparison
• —✅ Grad-CAM: Model interpretability with heatmaps
• —✅ Confusion Matrix: Per-class performance metrics
• —✅ Model Export: Download trained models (.pth)
• —✅ MLflow Tracking: Experiment versioning
• —✅ Tabular ML: Support for CSV/Excel data (6 models)
• —✅ Bilingual: English/German interface

GitHub Repository: kitsakisGk/AutoML-Forge

👤 Author

Kitsakis Giorgos

• —GitHub: @kitsakisGk
• —LinkedIn: georgios-kitsakis-gr

📄 License

MIT License