Import a simple image directory as dataset in Fiftyone

Question

I am new to Fifty one and would like to import a simple directory (containing .jpg files) into Fiftyone (I am using Windows and Python 3.10.10 in a Conda env). I followed the documentation instructions (https://docs.voxel51.com/user_guide/dataset_creation/index.html), however my dataset seems empty when opened in the app.

Here is my Python script :

import fiftyone as fo

DATASET_PATH = "\sandbox\images"

def load_dataset(path=DATASET_PATH, dataset_type=fo.types.ImageDirectory):
    dataset = fo.Dataset.from_dir(path, dataset_type)
    print(dataset)
    return dataset

def main():
    dataset = load_dataset()
    session = fo.launch_app(dataset, desktop=True)
    session.wait()

if __name__ == "__main__":
    main()

Here is the result of my import :

(fiftyone) C:\Windows>C:/Users/benoit.boidin/AppData/Local/anaconda3/envs/fiftyone/python.exe /sandbox/load_dataset.py
 100% |████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 0/0 [397.6us elapsed, ? remaining, ? samples/s] 
Name:        2024.04.10.13.13.22
Media type:  None
Num samples: 0
Persistent:  False
Tags:        []
Sample fields:
    id:       fiftyone.core.fields.ObjectIdField
    filepath: fiftyone.core.fields.StringField
    tags:     fiftyone.core.fields.ListField(fiftyone.core.fields.StringField)
    metadata: fiftyone.core.fields.EmbeddedDocumentField(fiftyone.core.metadata.Metadata)
Desktop App launched.

I tried several formats, using the webapp and desktop app, with large or small datasets. I expected the images to be displayed but everything work as if there was no files in my directory.

Answer 1

You want to use from_images_dir() here, not from_dir(). Here's the distinction:

• from_images_dir() assumes that the directory contains a bunch of image files
• from_dir() assumes there is a FiftyOne.Dataset that has been exported to that directory.