.NET Machine Learning Lab

A comprehensive collection of 20+ Machine Learning, Computer Vision, NLP, and Generative AI projects — all in C# and .NET

---

Overview

This repository demonstrates how to build real-world Machine Learning and AI applications using the .NET ecosystem. It covers a wide range of domains — from image classification and object detection to large language models and speech recognition — all implemented in C# with production-grade libraries.

---

Table of Contents

• Project Catalog
  • Computer Vision
  • Large Language Models & Generative AI
  • Natural Language Processing
  • Speech & OCR
  • Semantic Kernel
  • Utilities
• Tech Stack
• Prerequisites
• Getting Started
• License

---

Project Catalog

🔍 Computer Vision

Project	Description	Key Packages
ImageClassification	TensorFlow Inception model for ImageNet classification (1000 classes)	ML.NET, TensorFlow.NET
tf_image_classification	TensorFlow image scoring via ML.NET integration	ML.NET 3.0, TensorFlow.NET 2.16
mlnet_image_classification	Deep learning CNN training for concrete crack detection (SDNET2018 dataset)	ML.NET 4.0, ML.NET Vision
object_detection	Real-time object detection with Faster R-CNN — detects 80 COCO object categories	ONNX Runtime 1.16, ImageSharp
nsfw	NSFW content classifier using a fine-tuned Vision Transformer (ViT)	ONNX Runtime, ImageSharp
florence_2	Multi-task vision foundation model — OCR, captioning, object detection, segmentation	Florence2, ONNX Runtime 1.18
dalle3	DALL-E 3 text-to-image generation via Semantic Kernel	Semantic Kernel 1.19, SkiaSharp

🤖 Large Language Models & Generative AI

Project	Description	Key Packages
phi3	Interactive Phi-3-mini chatbot with local ONNX inference (4K context)	OnnxRuntimeGenAI 0.3
phi3vision	Phi-3 Vision 128K multimodal model — understands text + images	OnnxRuntimeGenAI, Spectre.Console
ai_toolkit	AI Toolkit inference with Phi-3-mini-128k (quantized INT4 ONNX)	OnnxRuntimeGenAI, DirectML, CUDA
bert_onnx	BERT large-uncased question-answering with ONNX Runtime	BERTTokenizers, ONNX Runtime, ML.NET
torchsharp	Sentiment analysis using TorchSharp (PyTorch bindings for .NET)	ML.NET TorchSharp 0.22

🌐 Natural Language Processing

Project	Description	Key Packages
cld_3	Language detection using Google's Compact Language Detector 3 (100+ languages)	Panlingo CLD3
fasttext	Language identification using Facebook's FastText (176+ languages)	Panlingo FastText

🎙️ Speech & OCR

Project	Description	Key Packages
whisper_net	ASP.NET Core Web API for speech-to-text using OpenAI Whisper	Whisper.net 1.7, NAudio
tesseract_dotnet	Optical Character Recognition with multi-language support	TesseractOCR 5.3, Spectre.Console

🧠 Semantic Kernel

A suite of projects demonstrating Microsoft Semantic Kernel — the SDK that integrates LLMs with conventional programming languages.

Project	Description	Key Packages
semantic_kernel/sk1	Chat with Azure OpenAI using prompts & plugins	Semantic Kernel 1.15
semantic_kernel/Chroma_app	RAG with Chroma vector database for semantic memory	SK Connectors.Chroma
semantic_kernel/native_functions	Extend LLMs with native C# plugins (e.g., MusicLibrary)	Semantic Kernel 1.16
semantic_kernel/ollamasharp	Run Phi-3 locally via Ollama + Semantic Kernel	OllamaSharp 2.0
semantic_kernel/phi3_sk	Phi-3 ONNX integrated with Semantic Kernel via custom connector	SK + OnnxRuntimeGenAI
semantic_kernel/planners	AI Planners — automatic plugin orchestration with Handlebars	SK Planners.Handlebars
semantic_kernel/dalle3	DALL-E 3 image generation through Semantic Kernel	Semantic Kernel 1.19

🛠️ Utilities

Project	Description	Key Packages
extract_images	Extract images from PDF, DOCX, PPTX, ODT, HTML, RTF documents	GroupDocs.Parser
mime	MIME type detection by inspecting file content (magic bytes)	Mime-Detective

---

Tech Stack

Category	Technologies
Runtime	.NET 8.0 (C# 12)
ML Frameworks	ML.NET, TorchSharp, TensorFlow.NET
Inference Engines	ONNX Runtime, OnnxRuntimeGenAI
LLMs	Phi-3, Phi-3 Vision, BERT, GPT (via Azure OpenAI)
Orchestration	Microsoft Semantic Kernel, Handlebars Planners
Vision Models	Florence-2, Faster R-CNN, ViT, Inception
Speech	Whisper.net (OpenAI Whisper)
OCR	Tesseract, Florence-2
Image Processing	SixLabors.ImageSharp, SkiaSharp
Vector Stores	Chroma DB
Web	ASP.NET Core, Kestrel
Containers	Docker (Whisper API, Ollama, Chroma)

---

Prerequisites

• .NET 8.0 SDK (or .NET Core 2.1 for legacy ImageClassification project)
• Visual Studio 2022 or VS Code with C# extension
• Visual C++ Redistributable (required for Tesseract, ONNX Runtime)

Model Downloads (project-specific):

Model	Download	Used By
Phi-3 Mini ONNX	huggingface-cli download microsoft/Phi-3-mini-4k-instruct-onnx --include cpu_and_mobile/cpu-int4-rtn-block-32-acc-level-4/* --local-dir .	phi3, phi3_sk, ai_toolkit
Phi-3 Vision ONNX	huggingface-cli download microsoft/Phi-3-vision-128k-instruct-onnx-cpu --include cpu-int4-rtn-block-32-acc-level-4/* --local-dir .	phi3vision
BERT Large ONNX	HuggingFace (convert to ONNX)	bert_onnx
Florence-2 ONNX	florence2-sharp	florence_2
FastText Models	lid.176.bin / fasttext217.bin	fasttext
Tesseract tessdata	tessdata	tesseract_dotnet
Whisper ggml-base	Auto-downloaded on first run	whisper_net
Faster R-CNN ONNX	ONNX Model Zoo	object_detection

Optional Docker Services:

# Ollama (for local LLM inference)
docker run -d --gpus=all -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama
ollama run phi3

# Chroma (for vector store / RAG)
docker run -it --rm -p 8000:8000/tcp chromadb/chroma:latest

API Keys (for cloud-based projects):

• Azure OpenAI API key — required for sk1, dalle3, planners

---

Getting Started

1. Clone the repository

   git clone https://github.com/your-username/DotNet_ML.git
   cd DotNet_ML

2. Navigate to any project folder

   cd phi3

3. Restore dependencies and run

   dotnet restore
   dotnet run

Each project is self-contained with its own .sln and .csproj. Refer to the project-specific section above for any required model downloads or configuration.

---

License

This project is licensed under the terms of the MIT License.