Repository for DreamScape: Denoising CLIP Embeddings with User Images for Improved Visualization Reconstruction.
Read writeup for full description of project or presentation for an abridged description. The vector database system is available in the file: DreamEmbeddings.ipynb.
Dataset for experiments 6.1 and EEG data available at https://github.com/bbaaii/DreamDiffusion/ and the DreamDiffusionST2 location model has a training file and weight file available.
Dataset and pipeline for experiments 6.2 and fMRI data available at https://github.com/MedARC-AI/fMRI-reconstruction-NSD/. Replace the available utils.py and Retreivals.ipynb file under src with the modified versions we provide. Our utils.py is available here and you can download Retrievals.ipynb from https://drive.google.com/file/d/1ZQ6gHPWcPWvgWbQcYWnNKux0GwOpaZdX/view?usp=sharing
The results of experiments 6.2 are available under the noisy_retrievals folder and can be recreated by running our Retrievals.ipynb.