RECANT Demonstration Tool
2. Classification task:
Select a classification task.
The models try to predict the depicted person’s Gender.
3. Results:
Click to show Results.
Nine different models were trained on the same images for each task, with different (sub)sets of crowd-worker annotations. The same input image (above) was passed through each of the nine models, resulting in the following outputs (possible outputs: Male, Female):
| Model | Model Description | Classification Decision |
|---|---|---|
| CFD Annotators | Model trained on the norming data provided with the CFD. | Female |
| All Annotators | Model trained using all the annotations for all images. | Female |
| Random | Model that simulates the case where annotators generate labels without considering the image content. | Male |
| Men | Model trained using all the annotations provided by male crowdworkers. | Female |
| Women | Model trained using all the annotations provided by female crowdworkers. | Female |
| Black | Model trained using all the annotations provided by Black crowdworkers. | Female |
| Asian | Model trained using all the annotations provided by Asian crowdworkers. | Female |
| White | Model trained using all the annotations provided by White crowdworkers. | Female |
| Latino | Model trained using all the annotations provided by Latino crowdworkers. | Female |