DualAfford: Learning Collaborative Visual Affordance for Dual-gripper Manipulation
Authors: Yan Zhao, Ruihai Wu, Zhehuan Chen, Yourong Zhang, Qingnan Fan, Kaichun Mo, Hao Dong
ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments prove the effectiveness and superiority of our method over baselines. |
| Researcher Affiliation | Collaboration | Yan Zhao1,6 Ruihai Wu1,6 Zhehuan Chen1 Yourong Zhang1 Qingnan Fan5 Kaichun Mo3,4 Hao Dong1,2,6 1CFCS, CS Dept., PKU 2AIIT, PKU 3Stanford University 4NVIDIA Research 5Tencent AI Lab 6BAAI |
| Pseudocode | No | The paper describes its method in text and figures, but does not provide pseudocode or a clearly labeled algorithm block. |
| Open Source Code | No | The paper provides a project page link (https://hyperplane-lab.github.io/Dual Afford) and mentions 'Please check our video for more results', but does not explicitly state that the source code for the methodology is openly released or provide a direct link to a code repository. |
| Open Datasets | Yes | We set up a benchmark for experiments using shapes from Part Net-Mobility dataset (Mo et al., 2019; Xiang et al., 2020) and Shape Net dataset (Chang et al., 2015). |
| Dataset Splits | No | The paper states 'We further split the training set into training shapes and testing shapes' and provides counts for these splits in Table 3, but does not explicitly mention or specify a validation dataset split. |
| Hardware Specification | Yes | It takes around 4.5-6 days per experiment on a single Nvidia 3090 GPU. |
| Software Dependencies | No | The paper mentions using 'Robot Operating System (ROS) (Quigley et al., 2009)' and 'Move It! (Chitta et al., 2012)' but does not specify their version numbers. It also refers to 'Adam optimizer (Kingma & Ba, 2014)' but this is an algorithm, not a specific software dependency with a version. |
| Experiment Setup | Yes | The hyper-parameters used to train the Perception Module are as following: 32 (batch size); 0.001 (learning rate of the Adam optimizer (Kingma & Ba, 2014)) for all three networks. ... We use the following hyper-parameters to train the RL policy: 16384 (buffer size); 512 (batch size); 0.0002 (learning rate of the Adam optimizer (Kingma & Ba, 2014) for both the policy and Q-value network); |