train_d_fine

Train D-FINE object detection models.

Following best practices recommended by the official repository, this algorithm utilizes the Objects365 pre-trained model as a foundation for fine-tuning, enabling optimized performance for custom object detection tasks.

🚀 Use with Ikomia API

1. Install Ikomia API

We strongly recommend using a virtual environment. If you're not sure where to start, we offer a tutorial here.

pip install ikomia

2. Create your workflow

from ikomia.dataprocess.workflow import Workflow

# Init your workflow
wf = Workflow()    

# Add dataset loader
coco = wf.add_task(name="dataset_coco")

coco.set_parameters({
    "json_file": "path/to/json/annotation/file",
    "image_folder": "path/to/image/folder",
    "task": "detection",
}) 

# Add training algorithm
train = wf.add_task(name="train_d_fine", auto_connect=True)

# Launch your training on your data
wf.run()

☀️ Use with Ikomia Studio

Ikomia Studio offers a friendly UI with the same features as the API.

• If you haven't started using Ikomia Studio yet, download and install it from this page.
• For additional guidance on getting started with Ikomia Studio, check out this blog post.

📝 Set algorithm parameters

• model_name (str) - default 'dfine_m': Name of the D-FINE pre-trained model on Objects365 (Best generalization). Other model available:
  • dfine_s
  • dfine_l
  • dfine_x
• batch_size (int) - default '8': Number of samples processed before the model is updated.
• epochs (int) - default '50': Number of complete passes through the training dataset.
• dataset_split_ratio (float) – default '0.9': Divide the dataset into train and evaluation sets ]0, 1[.
• input_size (int) - default '640': Size of the input image.
• weight_decay (float) - default '0.000125': Amount of weight decay, regularization method.
• workers (int) - default '0': Number of worker threads for data loading (per RANK if DDP).
• lr (float) - default '0.00025': Initial learning rate. Adjusting this value is crucial for the optimization process, influencing how rapidly model weights are updated.
• output_folder (str, optional): path to where the model will be saved.
• config_file (str, optional): path to the training config file .yaml. Using a config file allows you to set all the train settings available.

Parameters should be in strings format when added to the dictionary.

from ikomia.dataprocess.workflow import Workflow

# Init your workflow
wf = Workflow()    

# Add dataset loader
coco = wf.add_task(name="dataset_coco")

coco.set_parameters({
    "json_file": "path/to/json/annotation/file",
    "image_folder": "path/to/image/folder",
    "task": "detection",
}) 

# Add training algorithm
train = wf.add_task(name="train_d_fine", auto_connect=True)

train.set_parameters({
    "model_name": "dfine_m",
    "epochs": "100",
    "batch_size": "6",
    "input_size": "640",
    "dataset_split_ratio": "0.9",
    "workers": "0",  # Recommended to set to 0 if you are using Windows
    "weight_decay": "0.000125",
    "lr": "0.00025",
    "output_folder": "Path/To/Output/Folder", # Default folder : runs 
    "model_weight_file": "", # Optional
    "config_file": "Path/To/Config/file", # Optional 
    
})

# Launch your training on your data
wf.run()