You can refer to the implementation of datasets in our library. For example, we define ImageList as a base class on image classification task. Then we inherit ImageList class to implement specific datasets, such as Office31.
The most straightforward way to train on custom datasets is to mimic our implementation. However, this can be flexible
, such as the implementation of Food101. And you may need to modify the data loading code correspondingly.
You can refer to the implementation of datasets in our library. For example, we define ImageList as a base class on image classification task. Then we inherit ImageList class to implement specific datasets, such as Office31.
The most straightforward way to train on custom datasets is to mimic our implementation. However, this can be flexible , such as the implementation of Food101. And you may need to modify the data loading code correspondingly.