API Reference¶

This section provides detailed API documentation for all AutoTimm classes and functions.

API Organization¶

graph LR
    A[<b>AutoTimm API</b>] --> B[<b>Models</b>]
    A --> C[<b>Data</b>]
    A --> D[<b>Training</b>]
    A --> E[<b>Config</b>]

    B --> B1[ImageClassifier]
    B --> B2[ObjectDetector]
    B --> B3[SemanticSegmentor]
    B --> B4[InstanceSegmentor]
    B --> B5[YOLOXDetector]

    C --> C1[ImageDataModule]
    C --> C2[DetectionDataModule]
    C --> C3[SegmentationDataModule]
    C --> C4[InstanceSegDataModule]
    C --> C5[MultiLabelDataModule]

    D --> D1[AutoTrainer]
    D --> D2[Callbacks]
    D --> D3[TunerConfig]

    E --> E1[TransformConfig]
    E --> E2[MetricConfig]
    E --> E3[LoggerConfig]

    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 B1 fill:#1976D2,stroke:#1565C0
    style B2 fill:#1976D2,stroke:#1565C0
    style B3 fill:#1976D2,stroke:#1565C0
    style B4 fill:#1976D2,stroke:#1565C0
    style B5 fill:#1976D2,stroke:#1565C0
    style C1 fill:#1976D2,stroke:#1565C0
    style C2 fill:#1976D2,stroke:#1565C0
    style C3 fill:#1976D2,stroke:#1565C0
    style C4 fill:#1976D2,stroke:#1565C0
    style C5 fill:#1976D2,stroke:#1565C0
    style D1 fill:#1976D2,stroke:#1565C0
    style D2 fill:#1976D2,stroke:#1565C0
    style D3 fill:#1976D2,stroke:#1565C0
    style E1 fill:#1976D2,stroke:#1565C0
    style E2 fill:#1976D2,stroke:#1565C0
    style E3 fill:#1976D2,stroke:#1565C0

Core Classes¶

Task Models¶

Class	Description
ImageClassifier	End-to-end image classifier with timm backbone
ObjectDetector	FCOS-style anchor-free object detector
SemanticSegmentor	DeepLabV3+ / FCN semantic segmentation
InstanceSegmentor	Mask R-CNN style instance segmentation

Data Modules¶

Class	Description
ImageDataModule	Image classification (folder/built-in/CSV)
MultiLabelImageDataModule	Multi-label classification from CSV
DetectionDataModule	Object detection (COCO/CSV)
SegmentationDataModule	Semantic segmentation
InstanceSegmentationDataModule	Instance segmentation (COCO/CSV)

CSV Datasets¶

For direct CSV dataset usage without DataModules, see CSV Data Loading API:

Dataset	Task	Description
CSVImageDataset	Classification	Single-label from CSV
MultiLabelImageDataset	Multi-Label	Multi-label from CSV
CSVDetectionDataset	Detection	Bounding boxes from CSV
CSVInstanceDataset	Instance Seg	Masks from CSV

Training¶

Class	Description
AutoTrainer	Configured PyTorch Lightning Trainer

Configuration Classes¶

Class	Description
BackboneConfig	Timm backbone configuration
FeatureBackboneConfig	Feature extraction backbone configuration
TransformConfig	Unified transform configuration for models and data
MetricConfig	Single metric configuration
MetricManager	Multiple metrics manager
LoggingConfig	Enhanced logging options
LoggerConfig	Logger backend configuration
LoggerManager	Multiple loggers manager
TunerConfig	Auto-tuning configuration

Manager Classes API¶

MetricManager and LoggerManager share a consistent interface:

Method	MetricManager	LoggerManager
`len(manager)`	Number of configs	Number of loggers
`iter(manager)`	Iterate over configs	Iterate over loggers
`manager[i]`	Get config by index	Get logger by index
`manager.configs`	List of MetricConfig	List of LoggerConfig
Get by name	`get_metric_by_name(name)`	`get_logger_by_backend(name)`
Get config	`get_config_by_name(name)`	-

Heads¶

Head	Description
ClassificationHead	Simple classification head with dropout
DetectionHead	FCOS-style detection head (cls, bbox, centerness)
DeepLabV3PlusHead	DeepLabV3+ head with ASPP and decoder
FCNHead	Fully Convolutional Network head
MaskHead	Mask prediction head for instance segmentation
FPN	Feature Pyramid Network for multi-scale features
ASPP	Atrous Spatial Pyramid Pooling module

Loss Functions¶

Detection Losses¶

Loss	Description
FocalLoss	Focal loss for addressing class imbalance
GIoULoss	Generalized IoU loss for bounding box regression
CenternessLoss	Binary cross-entropy for centerness prediction
FCOSLoss	Combined FCOS detection loss

Segmentation Losses¶

Loss	Description
DiceLoss	Dice loss for handling class imbalance in segmentation
FocalLossPixelwise	Pixel-wise focal loss for segmentation
TverskyLoss	Generalized Dice with FP/FN control
MaskLoss	Binary cross-entropy for instance masks
CombinedSegmentationLoss	Combined CE + Dice loss

Logging¶

Symbol	Description
logger	Shared loguru logger instance for the entire library
log_table	Log a formatted ASCII table via loguru

Utility Functions¶

Function	Description
create_backbone	Create a headless timm model
create_feature_backbone	Create a feature extraction backbone
list_backbones	Search available timm models
get_feature_info	Get feature information from backbone
get_feature_channels	Extract feature channels for each level
get_feature_strides	Get stride information for FPN construction
resolve_backbone_data_config	Get model-specific preprocessing config
get_transforms_from_backbone	Create transforms with model normalization
create_inference_transform	Create simple inference transforms
seed_everything	Set random seeds for reproducibility
count_parameters	Count model parameters
list_optimizers	List available optimizers
list_schedulers	List available schedulers

Quick Links¶

Image Classification¶

from autotimm import (
    AutoTrainer,
    ImageClassifier,
    ImageDataModule,
    MetricConfig,
    MetricManager,
    TransformConfig,  # For preprocessing
)

Object Detection¶

from autotimm import (
    AutoTrainer,
    ObjectDetector,
    DetectionDataModule,
    FeatureBackboneConfig,
    MetricConfig,
)

Semantic Segmentation¶

from autotimm import (
    AutoTrainer,
    SemanticSegmentor,
    SegmentationDataModule,
    MetricConfig,
)

# Using import aliases
from autotimm.task import SemanticSegmentor
from autotimm.loss import DiceLoss, CombinedSegmentationLoss
from autotimm.head import DeepLabV3PlusHead, FCNHead

Instance Segmentation¶

from autotimm import (
    AutoTrainer,
    InstanceSegmentor,
    InstanceSegmentationDataModule,
    MetricConfig,
)

# Using import aliases
from autotimm.task import InstanceSegmentor
from autotimm.loss import MaskLoss
from autotimm.head import MaskHead

With Logging¶

from autotimm import (
    LoggerConfig,
    LoggerManager,
    LoggingConfig,
)

With Auto-Tuning¶

from autotimm import TunerConfig

Backbone Utilities¶

from autotimm import (
    BackboneConfig,
    FeatureBackboneConfig,
    create_backbone,
    create_feature_backbone,
    get_feature_channels,
    get_feature_info,
    get_feature_strides,
    list_backbones,
)

Logging¶

from autotimm import logger
from autotimm.core.logging import logger, log_table

Utilities¶

from autotimm import (
    count_parameters,
    list_optimizers,
    list_schedulers,
)

Module Structure¶

autotimm/
├── __init__.py                # Public API exports
├── backbone.py                # BackboneConfig, FeatureBackboneConfig, create_backbone,
│                              # create_feature_backbone, get_feature_*, list_backbones
├── data/
│   ├── datamodule.py          # ImageDataModule
│   ├── dataset.py             # ImageFolderCV2
│   ├── transforms.py          # Transform presets
│   ├── transform_config.py    # TransformConfig
│   ├── timm_transforms.py     # Model-specific transforms
│   ├── detection_datamodule.py # DetectionDataModule
│   ├── detection_dataset.py   # DetectionDataset
│   ├── detection_transforms.py # Detection augmentations
│   ├── segmentation_dataset.py # SemanticSegmentationDataset
│   ├── segmentation_datamodule.py # SegmentationDataModule
│   ├── instance_dataset.py    # COCOInstanceDataset
│   ├── instance_datamodule.py # InstanceSegmentationDataModule
│   └── segmentation_transforms.py # Segmentation augmentations
├── heads.py                   # ClassificationHead, DetectionHead, FPN
│                              # DeepLabV3PlusHead, FCNHead, MaskHead, ASPP
├── loggers.py                 # LoggerConfig, LoggerManager
├── losses/
│   ├── detection.py           # FocalLoss, GIoULoss, CenternessLoss, FCOSLoss
│   └── segmentation.py        # DiceLoss, FocalLossPixelwise, TverskyLoss,
│                              # MaskLoss, CombinedSegmentationLoss
├── metrics.py                 # MetricConfig, MetricManager, LoggingConfig
├── tasks/
│   ├── classification.py      # ImageClassifier
│   ├── object_detection.py    # ObjectDetector
│   ├── semantic_segmentation.py # SemanticSegmentor
│   ├── instance_segmentation.py # InstanceSegmentor
│   └── preprocessing_mixin.py # PreprocessingMixin
├── logging.py                 # Central loguru-based logger, log_table
├── trainer.py                 # AutoTrainer, TunerConfig
└── utils.py                   # Utility functions