Skip to content

README.md

Latest commit

 

History

History
66 lines (41 loc) · 2.94 KB

README.md

File metadata and controls

66 lines (41 loc) · 2.94 KB
Raw

Dataset Setup

This repository uses the following datasets: M3ED, DSEC and MVSEC. If you download these datasets in some location, you can use the scripts provided in scripts/setup/setup_<dataset_name>.sh to create symbolic links to these datasets in the data/ folder. These scripts also create the required 50kHz_timestamps.npy files.

M3ED

The M3ED dataset is present here. To download the necessary sequences for car, spot, and falcon, use the script download_m3ed.py. Run the following scripts to download the sequences:

# Car urban driving sequences
python3 scripts/download/download_m3ed.py --vehicle car --environment urban --output_dir /path/to/M3ED/

# Spot indoor and outdoor sequences
python3 scripts/download/download_m3ed.py --vehicle spot --environment indoor --output_dir /path/to/M3ED/
python3 scripts/download/download_m3ed.py --vehicle spot --environment outdoor --output_dir /path/to/M3ED/

# Falcon indoor and outdoor sequences
python3 scripts/download/download_m3ed.py --vehicle falcon --environment indoor --output_dir /path/to/M3ED/
python3 scripts/download/download_m3ed.py --vehicle falcon --environment outdoor --output_dir /path/to/M3ED/

After downloading the necessary sequences, run the following commands to set up the symbolic links in data/ and 50kHz_timestamps.npy files:

bash scripts/setup_m3ed_car.sh /path/to/M3ED/
bash scripts/setup_m3ed_spot.sh /path/to/M3ED/
bash scripts/setup_m3ed_falcon.sh /path/to/M3ED/

Note: If you followed the instructions above to download the car sequences, you can train an F3 model on the car urban daytime driving sequences using the command shown in the Training an F3 section of the README. Setting up F3 training for other M3ED sequences is straightforward -- choose the appropriate configuration file from confs/ff/trainoptions/ and enjoy!

DSEC

Download the DSEC dataset from here. After downloading and extracting the dataset, run the following command to set up the symbolic links and timestamps.npy files:

bash scripts/setup_dsec.sh /path/to/DSEC/ data/ 1 # Sets up DSEC train split in data/
bash scripts/setup_dsec.sh /path/to/DSEC/ data/ 0 # Sets up DSEC test split in data/

MVSEC

Download the MVSEC dataset from here. After downloading the .hdf5 files, run:

python3 scripts/process_mvsec.py /path/to/MVSEC/<XYZ>.hdf5 /path/to/MVSEC/<XYZ>_processed.hdf5

for the sequences outdoor_day1, outdoor_day2, and outdoor_night1. Then run:

bash scripts/setup_mvsec.sh /path/to/MVSEC/

TODO

  • timestamps 50KHz file requirements have been removed. Update instructions accordingly.

  • Add instructions for downloading tartanairv2 dataset.