Mar 11, 2022 · We demonstrate gains of the consistency and spatial consistency losses over the binary cross-entropy baseline, and over competing methods, on MS ...

In such setting, a single-labeled dataset can be thought of as a weakly-labeled multi-label classification dataset, with a single positive annotation per image.

We train multi-label classifiers from datasets where each image is annotated with a single positive label only. As the presence of all other classes is unknown, ...

We see that CL and SCL are able to benefit more from the crop data- augmentation, compared to AN. This is consistent with our intuition that the crop data- ...

Sep 7, 2024 · We demonstrate gains of the consistency and spatial consistency losses over the binary cross-entropy baseline, and over competing methods, on MS ...

MLZSL can be viewed as an extension of both multi-label learning and zero-shot learning. In multi-label learning, it strives to recognize all the categories ...

Mar 11, 2022 · We introduce a spatial consistency loss (SCL) to further mitigate multi-object label noise in single-labeled datasets, and address the label ...

This is a collection of papers and code for single positive multi-label learning (SPML), an interesting and challenging variant of multi-label learning.

We show that adding a consistency loss, ensuring that the predictions of the network are consistent over consecutive training epochs, is a simple yet effective ...

Spatial Consistency Loss for Training Multi-Label Classifiers from Single-Label Annotations ... Consistency for Multi-Label Learning with Single Positive Labels.