Builtins
Computation Modules
OneLabeler builds in the implementation of a variety of interface and algorithm modules at developers' disposal. These modules can be composed into data labeling tools without requiring textual programming.
See the full list of built-in modules
| Module | Type | interface/algorithm |
|---|---|---|
| bitmap image | interactive labeling - data type | interface |
| text | interactive labeling - data type | interface |
| video | interactive labeling - data type | interface |
| audio | interactive labeling - data type | interface |
| point cloud | interactive labeling - data type | interface |
| vector image | interactive labeling - data type | interface |
| interactive labeling - data type | interface | |
| website | interactive labeling - data type | interface |
| classification | interactive labeling - label task | interface |
| multi-label classification | interactive labeling - label task | interface |
| freeform text annotation | interactive labeling - label task | interface |
| object detection | interactive labeling - label task | interface |
| 2D segmentation | interactive labeling - label task | interface |
| 3D segmentation | interactive labeling - label task | interface |
| span tagging | interactive labeling - label task | interface |
| span relation | interactive labeling - label task | interface |
| temporal segmentation | interactive labeling - label task | interface |
| single object display | interactive labeling | interface |
| grid matrix | interactive labeling | interface |
| cluster centroids | data object selection | algorithm |
| cluster index | data object selection | algorithm |
| dense areas | data object selection | algorithm |
| entropy | data object selection | algorithm |
| entropy diversity | data object selection | algorithm |
| entropy diversity density | data object selection | algorithm |
| least confident | data object selection | algorithm |
| smallest margin | data object selection | algorithm |
| interactive projection | data object selection | interface |
| COCO-SSD | default labeling | algorithm |
| DeepLab | default labeling | algorithm |
| MobileNet | default labeling | algorithm |
| POS tagging | default labeling | algorithm |
| general-purpose model prediction | default labeling | algorithm |
| image bag of words | feature extraction | algorithm |
| image linear discriminant analysis | feature extraction | algorithm |
| image singular value decomposition | feature extraction | algorithm |
| text non-negative matrix factorization | feature extraction | algorithm |
| MobileNet extractor | feature extraction | algorithm |
| retrain decision tree | model training | algorithm |
| retrain SVM | model training | algorithm |
| retrain logistic regression | model training | algorithm |
| retrain RBN | model training | algorithm |
| online graph-based label propagation | model training | algorithm |
| one by one checking | quality assurance | interface |
| labeled rate | stoppage analysis | algorithm |
| editable label category table | label ideation | interface |
Labeling Tool Templates
OneLabeler builds in a collection of template data labeling tools. These templates can be used directly, or used as boilerplate for creating new data labeling tools.
The list of built-in templates:
- audio classification
- audio temporal segmentation
- image classification
- mixed-initiative image classification
- image segmentation
- point cloud classification
- point cloud segmentation
- text classification
- text named entity recognition
- video classification
- video temporal segmentation
- webpage classification