Multi-label Classification

Multi-label classification is a generalization of multiclass classification. The main goal of the task is to predict a set of labels per image. Each image may belong to more than one class and may belong to none of them at all.

We solve this problem by optimizing small binary classification sub-tasks aimed to predict whether or not the specific category from the corpus is presented on the given image.

For supervised learning we use the following algorithms components:

• Learning rate schedule: ReduceLROnPlateau. It is a common learning rate scheduler that tends to work well on average for this task on a variety of different datasets.

• Loss function: We use Asymmetric Angular Margin Loss. We can formulate this loss as follows: \(L_j (cos\Theta_j,y) = \frac{k}{s}y p_-^{\gamma^-}\log{p_+} + \frac{1-k}{s}(1-y)p_+^{\gamma^+}\log{p_-}\), where \(s\) is a scale parameter, \(m\) is an angular margin, \(k\) is negative-positive weighting coefficient, \(\gamma^+\) and \(\gamma^-\) are weighting parameters. For further information about loss function, ablation studies, and experiments, please refer to our dedicated paper.

• Additionally, we use the early stopping to add adaptability and prevent overfitting.

Dataset Format

The format should be converted in our internal representation.

Note

Names of the annotations files and overall dataset structure should be the same as above example. You need to convert train and validation sets separately.

Please, refer to our dedicated tutorial for more information how to train, validate and optimize classification models.

Models

We use the same models as for Multi-class classification. Please, refer: Classification Models.

To see which models are available for the task, the following command can be executed:

(otx) ...$ otx find --task MULTI_LABEL_CLS

In the table below the mAP metrics on some academic datasets using our supervised pipeline are presented. The results were obtained on our recipes without any changes (including input resolution, which is 224x224 for all recipes). We trained each model with single Nvidia GeForce RTX3090.

Model name	Pascal-VOC 2007	COCO 2014	Aerial Maritime	Mean mAP
MobileNet-V3-large-1x	86.14	67.94	69.61	74.56
EfficientNet-B0	86.07	67.87	73.83	75.92
EfficientNet-V2-S	91.91	77.28	71.52	80.24