AutoDL automates machine learning tasks enabling you to easily achieve strong predictive performance in your applications. With just a few lines of code, you can train and deploy high-accuracy machine learning and deep learning models on image.

• automl benchmark

• benchmark repo

• [NAS benchmark]

This tutorial demonstrates how to use AutoDL with your own custom datasets. As an example, we use a dataset from Kaggle to show the required steps to format image data properly for AutoDL.

After completing this step, you will have the following directory structure on your machine:

   Your_Dataset/
    ├──train/
        ├── class1/
        ├── class2/
        ├── class3/
        ├── ...
    ├──test/
        ├── class1/
        ├── class2/
        ├── class3/
        ├── ...

Here Your_Dataset is a folder containing the raw images categorized into classes. For example, subfolder class1 contains all images that belong to the first class, class2 contains all images belonging to the second class, etc.

We generally recommend at least 100 training images per class for reasonable classification performance, but this might depend on the type of images in your specific use-case.

Under each class, the following image formats are supported when training your model:

- JPG
- JPEG
- PNG

In the same dataset, all the images should be in the same format. Note that in image classification, we do not require that all images have the same resolution.

You will need to organize your dataset into the above directory structure before using AutoDL.

Sometimes dataset needs additional data preprocessing by Script data_processing.

  data
    ├──XXXX/images_all
    ├         ├── img1.jpg
    ├         ├── img2.jpg
    ├──XXXX/test
    ├         ├── ...

python data_processing.py --dataset <aerial\dog\> --data-dir data

Finally, we have the desired directory structure under ./data/XXXX/train/, which in this case looks as follows:

  data
    ├──XXXX/train
    ├         ├── classA
    ├         ├── classb
    ├         ├── ...
    ├──XXXX/test
    ├         ├── ...
    ├
    ├
    ├──ZZZZ/train
    ├         ├── classA
    ├         ├── classb
    ├         ├── ...
    ├──ZZZZ/test
              ├── ...

python data_processing.py --dataset <aerial\dog\> --data-dir data

Sometimes dataset needs additional data_split by Script data_split.

dataset__name
    ├──train
        ├──split/train
        ├         ├── classA
        ├         ├── classb
        ├         ├── ...
        ├──split/test
        ├         ├── classA
        ├         ├── classb
        ├         ├── ...
    ├──test
        ├── img1.jpg
        ├── img2.jpg
        ├── ...

python data_split.py --data-dir /data/AutoML_compete/Store-type-recognition/

Now that we have a Dataset object, we can use AutoGluon's default configuration to obtain an image classification model using the fit function.

Run benchmark.py script with different dataset:

python benchmark.py \
    --data_path /media/robin/DATA/datatsets/image_data/dog-breed-identification \
    --output_path /home/robin/jianzh/automl/autodl/benchmark \
    --dataset dog-breed-identification \
    --train_framework autogluon

python benchmark.py \
    --data_path /media/robin/DATA/datatsets/image_data/hymenoptera/images/train \
    --dataset hymenoptera \
    --output_path /home/robin/jianzh/automl/autodl/benchmark \
    --train_framework autokeras

Bag of tricks are used on image classification dataset.

Customize parameter configuration according your data as follow:

lr_config = ag.space.Dict(
            lr_mode='cosine',
            lr_decay=0.1,
            lr_decay_period=0,
            lr_decay_epoch='40,80',
            warmup_lr=0.0,
            warmup_epochs=5)

tricks = ag.space.Dict(
            last_gamma=True,
            use_pretrained=True,
            use_se=False,
            mixup=False,
            mixup_alpha=0.2,
            mixup_off_epoch=0,
            label_smoothing=True,
            no_wd=True,
            teacher_name=None,
            temperature=20.0,
            hard_weight=0.5,
            batch_norm=False,
            use_gn=False)