In this paper, we study the task of embodied interactive learning for object detection. Given a set of environments (and some labeling budget), our goal is to learn an object detector by having an agent select what data to obtain labels for. How should an exploration policy decide which trajectory should be labeled? One possibility is to use a trained object detector's failure cases as an external reward. However, training RL policies requires millions of labelled samples, hence making any supervision infeasible. Instead, we explore a self-supervised approach for training our exploration policy by introducing a notion of semantic curiosity. Our semantic curiosity policy is based on a simple observation -- the detection outputs should be consistent. Therefore, our semantic curiosity rewards trajectories with inconsistent labeling behavior and encourages the exploration policy to explore such areas. The exploration policy trained via semantic curiosity generalizes to novel scenes and helps train an object detector that outperforms baselines trained with other possible alternatives such as random exploration, prediction-error curiosity and coverage-maximizing exploration.

We use semantic curiosity to learn an exploration policy on set of the training environments. The exploration policy is learned by projecting segmentation masks on the top-down view to create semantic maps. The entropy of semantic map defines the inconsistency of the object detection module. The learned exploration policy is then used to generate training data for object detection/segmentation module. The labeled data is then used to finetune and evaluate the object detection/segmentation.

@inproceedings{chaplot2020semantic,
  title={Semantic Curiosity for Active Visual Learning},
  author={Chaplot, Devendra Singh and Jiang, Helen and Gupta, Saurabh and 
          Gupta, Abhinav},
  booktitle={ECCV}
  year={2020}}