Ali Farhadi, Roozbeh Mottaghi, Luca Weihs, Kuo-Hao Zeng
We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. The repositories linked to it are listed below.
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
Figure 1: Visual navigation may require interactions that go beyond moving forward/backward, and turning left/right. For example, the agent in the top row needs to push the chair out of its way to reach the target. Interactive navigation entails deeper understanding of the outcome of agents actions on objects in the scene. In this paper, we introduce Neural Interaction Engine (NIE) to explicitly predict the effect of actions on objects poses. By integrating NIE with our policy network we show that we can perform long-horizon planning while predicting the outcome of the actions. We evaluate NIE for visual navigation where the path to the goal is obstructed, and moving objects to specific locations in the scene and show major improvements over state of the art in these tasks.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2104.14040")
get_code_for_paper("2104.14040")
have("2104.14040")
The run record, dated, one paper per request, free:
curl https://syntology.ai/api/ran/2104.14040.json
A badge for a README (the split and the date, never a ratio):
[](https://syntology.ai/paper/2104.14040)
Connect an agent — have() is free.