Conference

ICCC 2020

Authors

• Medhini Narasimhan
• Erik Wijmans
• Xinlei Chen
• Trevor Darrell
• Dhruv Batra
• Devi Parikh
• Amanpreet Singh

Contributions

• Introduced a novel learning-based approach for room navigation via a modal prediction of semantic maps. The agent learns architectural and stylistic regularities in houses to predict regions beyond its field of view.
• Through carefully designed ablations, we show that our model trained to predict semantic maps as intermediate representations achieves better performance on unseen environments compared to a baseline which doesn’t explicitly generate semantic top-down maps.
• To evaluate our approach, we introduce the room navigation task and dataset in the Habitat platform.

Approach

• Room Navigation Task : Using sensors.
• Room Navigation using Amodal semantic maps : Based on general ideas of house designs, Uses Seq2Seq network

• Interesting details :

  • Map generation : ResNet50
  • Point Prediction : ResNet50
  • Point Navigation Policy : ResNeXt

Dataset

• Matterport 3D