This is the pytorch implementation of CVPR 2024 paper: Instance-aware Exploration-Verification-Exploitation for Instance ImageGoal Navigation (IEVE).
Inspired by the human behavior of “getting closer to confirm” when recognizing distant objects, we formulate the task of determining whether an object matches the one in the goal image as a sequential decision problem. In addition, we design a novel matching function that relies not only on the current observation and goal image but also on the Euclidean distance between the agent and the potential target. We categorize the targets into confirmed target, potential target, and no-target (exploration), and allow the agent to actively choose among these three targets.
pip install -e 3rdparty/habitat-lab
pip install -r requirements.txt
data
To test in the val set, run:
python main.py
agents\instance_exp.py
to see the Switch Policy@inproceedings{lei2024instance,
title={Instance-aware Exploration-Verification-Exploitation for Instance ImageGoal Navigation},
author={Lei, Xiaohan and Wang, Min and Zhou, Wengang and Li, Li and Li, Houqiang},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={16329--16339},
year={2024}
}