HyperAI

Multi Label Image Recognition With Partial

Multi-label image recognition with partial labels (MLR-PL) refers to the task of training a multi-label recognition model to reduce annotation costs in scenarios where some labels are known while others are unknown in multi-label images. The goal is to train the model using a partially labeled dataset and evaluate its performance when the proportion of known labels ranges from 10% to 90%, thereby enhancing the model's practicality and cost-effectiveness.