Event-based vision utility scripts.
- preprocess.py: Preprocessing of frames. Consists of resizing, downsampling and cropping, all in batches.
- bag2hdf5.py: Converts a .rosbag file containing events into a .hdf5. MVSEC dataset is taken as a guideline for the data organization.
- bag2img.py: Extracts image_raw data from a rosbag file.
- img2hdf5.py: Writes grayscale image data read from a directory into an existing hdf5 file.
- flow2hdf5.py: Writes .flo files read from a directory into a new hdf5 file, with their corresponding timestamps to the output of bag2hdf5.py.
- metrics.py Includes some optical flow evaluation metrics.
- bin_events.py Using count_data output of Spike-FlowNet, generates accumulated binary event images.
- auto.bash: Runs all necessary scripts for:
- Pre-processing (resizing, downsampling etc.) high framerate images and their flow files
- Generating events in .rosbag format via ESIM using processed high framerate images & flows
- Converting .rosbag dataset into .hdf5, taking MVSEC dataset as a guideline
- Split encoding of data in hdf5 file (part of Spike-FlowNet)
- Before running auto.bash, roscore should be run beforehand, and paths should be changed accordingly.