MicroExpressions

Catch people micro-expressions using an RGB camera and an Event camera. Then try to classify them: which is the best ?

Frame coding

Event camera capture a stream of events. This stream needs to be 'converted' in frames, in order to deal with neural network processing. Different codings are possible, here we work with Temporal Binary Encoding. Please look at link to get more information about the coding. Here we install only the necessary files. The prophesee-automotive-dataset-toolbox has been downloaded from link and added to the sys path so it can be used as a project module. Folder frame_coding contains encoders.py and tbe.py from link and the files related to the conversion of the video streams of the dataset in frames.

Proposed folder hierarchy

dataset                                    
│
├── event_videos
│    ├── original
│        ├── user_01
│            ├── user_01-timestamp.raw
│            ├── user_01-timestamp.raw
│            ├── user_01-timestamp.raw
│            ├── user_01-timestamp.raw
│            └── ...
│    ├── cut
│        ├── user_01
│            ├── user_01-timestamp.raw
│            ├── user_01-timestamp.raw
│            ├── user_01-timestamp.raw
│            ├── user_01-timestamp.raw
│            └── ...
│ 
├── gopro_videos
│   ├── user_01
│   ├── user_02
│   └── ...
│
├── dat_dataset
│   └── ...
├── frame_dataset
│   └── ...
├── csv_dataset  # uno per utente
│   └── 
└──