Controllable Mind Visual Diffusion Model
Authors: Bohan Zeng, Shanglin Li, Xuhui Liu, Sicheng Gao, Xiaolong Jiang, Xu Tang, Yao Hu, Jianzhuang Liu, Baochang Zhang
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Through extensive experimentation, we demonstrate that CMVDM outperforms existing state-of-the-art methods both qualitatively and quantitatively. |
| Researcher Affiliation | Collaboration | Bohan Zeng1*, Shanglin Li1*, Xuhui Liu1, Sicheng Gao1 Xiaolong Jiang3, Xu Tang3, Yao Hu3, Jianzhuang Liu4, Baochang Zhang1,2,5 1Institute of Artificial Intelligence, Hangzhou Research Institute, Beihang University, China 2Nanchang Institute of Technology, Nanchang, China 3Xiaohongshu Inc 4Shenzhen Institute of Advanced Technology, Shenzhen, China 5 Zhongguancun Laboratory, Beijing, China |
| Pseudocode | No | The paper describes its methodology using text and diagrams (Fig. 2) but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our code is available at https://github.com/zengbohan0217/CMVDM. |
| Open Datasets | Yes | In this study, we employ two public datasets with paired f MRI signals and images: Generic Object Decoding (GOD) dataset (Horikawa and Kamitani 2017), and Brain, Object, Landscape Dataset (BOLD5000) (Chang et al. 2019). |
| Dataset Splits | No | The GOD dataset... with 50 images designated for testing. ...BOLD5000... resulting in 4803 f MRI-image pairs for training and 113 for testing. |
| Hardware Specification | Yes | We adopt 1 A100-SXM4-40GB GPU for the training of Efmri and the control model, and 1 V100SXM2-32GB GPU for Dslh training. |
| Software Dependencies | No | Both Efmri and the control model are trained by the Adam W (Loshchilov and Hutter 2017) with β = (0.9, 0.999) and eps = 1e 8 for 500 epochs. Dslh is optimized using Adam (Kingma and Ba 2015) with a learning rate of 5e 3 and β = (0.5, 0.99) for 150 epochs. |
| Experiment Setup | Yes | Both Efmri and the control model are trained by the Adam W (Loshchilov and Hutter 2017) with β = (0.9, 0.999) and eps = 1e 8 for 500 epochs. Dslh is optimized using Adam (Kingma and Ba 2015) with a learning rate of 5e 3 and β = (0.5, 0.99) for 150 epochs. |