Refer to the MMFI Webpage for instructions to download the MMFi Dataset Process the data by placing all the data for a particular sample (subject performing an action) into a single .pickle file with the keys following the format in ./shared_files/PickleDataset.py.

After obtaining the dataset, please change all occurrences of /mnt/ssd_8t/jason/MMFI_Dataset/ with the correct dataset path via find and replace

Please refer to the following subfolders for instructions on how to run the training and evaluation

• MMFI_Code_2_FT: Evaluations after finetuning on two subjects
• MMFI_Code_4_FT: Evaluations after finetuning on four subjects
• MMFI_Code_8_FT: Evaluations after finetuning on eight subjects
• MMFI_Code_Diff_Envs: Evaluations the impact of domain shift across pretrain and finetune
• MMFI_Code_Label_Binding: Evaluate label binding results
• MMFI_Code_Random_Binding: Randomly generate the paired dataset for MMBind
• MMFI_Code_Weighted: The weighted contrastive learning results and baselines that we show in Table 3
• MMFI_Code_WiFi_Binding: Performing data binding with the non-descriptive WiFi modality