Fifty-one
Features Of fiftyone
pip install fiftyone
1
2
3
4
5
6
7
8
9
10
11
12
import fiftyone as fo
dataset = fo.Dataset("my_dataset")
sample = fo.Sample(filepath="/path/to/image.png")
sample.tags.append("train")
sample["custom_field"] = 51
dataset.add_sample(sample)
view = dataset.match_tags("train").sort_by("custom_field").limit(10)
for sample in view:
print(sample)
- 🗃️ Dataset: Think of it as a container for your data, composed of individual Sample objects.
- 📌 Sample: Represents wøa single data point (e.g., an image). Stores essential info like filepaths and custom fields.
- Each sample has an ID, file path, tags, label, field
- 🧰 Custom Fields: Add flexibility! Store anything from basic types (numbers, strings) to complex labels (bounding boxes, segmentations).
-
🪄 Automatic Media Type Detection: FiftyOne infers if a sample is an image, video, etc., based on its filepath.
-
The Dataset is backed by a MongoDB database
- are there 3D point cloud features?
- Clip model: reduce what?
- What is data curation:
- Search, filter by text.
- find similar images
Correspondent: Antonio Rueda Toicen
