[Master's Thesis] Improving Instruction Generation for Vision-Language Navigation by Reward Designing
A new method to enhance the performance of language generation model for vision-language navigation by using REINFORCE with multi-modal rewards
2021
[Accepted by AI-HRI, 2021] Natural Language for Human-Robot Collaboration：Problems Beyond Language Grounding
A blue sky paper points out challenges and opportunities in human-robot interaction, in the context of vision-language navigation
[Accepted by 3DV, 2021] GASCN: Graph Attention Shape Completion Network
A shape completion network utilizes graph attention and surface normal
2020
A Study of Policy Gradient Methods for Seq2Seq Model in Neural Machine Translation
Vanilla policy gradient methods could solve issues in classical MLE training, but how effective are they?
Feature-Based Stereo Visual Odometry
We compared the performance of visual SLAM packages. As for my task, I built the visual-odometry pipeline from scratch and compared it with SOFT-odometry.