User Guide¶

Comprehensive guides for using AutoTimm across all computer vision tasks — from data loading to deployment.

Guide Organization¶

graph LR
    A[<b>User Guide</b>] --> B[<b>Data Loading</b><br/>Datasets, transforms,<br/>presets]
    A --> C[<b>Models</b><br/>Classifier, detector,<br/>segmentor]
    A --> D[<b>Training</b><br/>AutoTrainer, losses,<br/>customization]
    A --> E[<b>Evaluation</b><br/>Metrics, benchmarks,<br/>selection]
    A --> F[<b>Interpretation</b><br/>GradCAM, attention,<br/>visualization]
    A --> G[<b>Integration</b><br/>HuggingFace Hub,<br/>Transformers]
    A --> H[<b>Deployment</b><br/>TorchScript, C++,<br/>mobile]
    A --> I[<b>Inference</b><br/>Classification,<br/>detection, segmentation]

    style A fill:#1565C0,stroke:#0D47A1
    style B fill:#1976D2,stroke:#1565C0
    style C fill:#1565C0,stroke:#0D47A1
    style D fill:#1976D2,stroke:#1565C0
    style E fill:#1565C0,stroke:#0D47A1
    style F fill:#1976D2,stroke:#1565C0
    style G fill:#1565C0,stroke:#0D47A1
    style H fill:#1976D2,stroke:#1565C0
    style I fill:#1565C0,stroke:#0D47A1

Browse by Category¶

Data Loading¶

Prepare and load data for any computer vision task.

• Overview - Data module overview and architecture
• Image Classification Data - ImageDataModule for classification tasks
• Multi-Label Classification Data - Multi-label image datasets
• Object Detection Data - Detection dataset formats (COCO, VOC)
• Segmentation Data - Semantic segmentation masks
• CSV Data Loading - Load datasets from CSV files
• Transforms - Augmentation and preprocessing pipelines
• Preset Manager - Pre-configured transform presets

Models¶

Task-specific model architectures built on timm.

• Overview - Model architecture overview
• ImageClassifier - Image classification with any timm backbone
• ObjectDetector - Object detection with FCOS
• SemanticSegmentor - Semantic segmentation with DeepLabV3+/FCN
• InstanceSegmentor - Instance segmentation with Mask R-CNN
• YOLOXDetector - Real-time detection with YOLOX

Training¶

Configure and run training with AutoTrainer.

• Overview - AutoTrainer setup and configuration
• Dataset Preparation - Preparing data for training
• Loss Comparison - Choosing the right loss function
• Advanced Customization - Custom callbacks, schedulers, and more

Evaluation¶

Measure and compare model performance.

• Metrics - MetricConfig and MetricManager
• Metric Selection - Choosing metrics per task
• Benchmarks - Performance benchmarks and comparisons

Interpretation¶

Understand what your models learn.

• Overview - Interpretation tools overview
• Interpretation Methods - GradCAM, attention, and more
• Quality Metrics - Quantitative interpretation metrics
• Interactive Visualizations - Plotly-based interactive views
• Performance Optimization - Efficient interpretation at scale
• Feature Visualization - Visualize learned features
• Training Callbacks - Log interpretations during training
• Task-Specific Adapters - Adapters for detection and segmentation

Integration¶

Connect with HuggingFace ecosystem.

• Comparison Overview - Compare integration approaches
• HF Hub Integration - Push/pull models from Hub
• HF Transformers Integration - Use Transformers models

Deployment¶

Deploy models to production.

• Production Deployment - Full production deployment guide
• TorchScript Export - Export models to TorchScript
• C++ Deployment - Deploy with LibTorch in C++
• Mobile Deployment - Deploy on iOS and Android

Inference¶

Run predictions with trained models.

• Overview - Inference workflow overview
• Classification Inference - Classify images
• Object Detection Inference - Detect objects
• Segmentation Inference - Segment images
• Model Export - Export for optimized inference

Guides¶

Best practices and reference guides.

• Reproducibility - Achieve reproducible results
• Logging - Configure experiment logging
• Migration Guide - Migrate from other frameworks
• YOLOX Quick Reference - YOLOX cheat sheet