MEVDT/							Root directory: contains all data and labels for MEVDT dataset
├── sequences/						Contains images and sequence-long event streams, organized into training and testing datasets
│ ├── train/
│ │ ├── Scene_A/
│ │ │ └── <sequence_name>/				Directory named after the timestamp of its first sample; holds related images and event data
│ │ │ ├── <image_timestamp>.png 			Image file named by its exact timestamp of capture
│ │ │ └── <sequence_name>_events.csv 			Event stream CSV file listing events captured throughout the sequence
│ │ └── Scene_B/ 					Structure follows the same format as Scene_A/
│ │ │ └── ...
│ └── test/ 						Testing data, organized similarly to the training data in train/
| └── ...
|
├── labels/ 						Ground truth data for object tracking and detection tasks
│ ├── tracking_labels/ 					Object tracking labels
│ │ ├── train/
│ │ │ └── Scene_A/
│ │ │ └── <sequence_name>/ 				Directory named after the timestamp of its first sample, containing ground truth in various formats
│ │ │ ├── <sequence_name>-coco.json 			Ground truth in COCO JSON format
│ │ │ ├── <sequence_name>-custom<tracking_rate>.txt 	Custom ground truth format for tracking at specified rates {24, 48, 96, 192, 384 Hz}
│ │ │ └── <sequence_name>-mot<tracking_rate>.txt 	MOT challenge format ground truth for specified tracking rates {24, 48, 96, 192, 384 Hz}
│ │ └── ... 						Repeat structure for additional scenes and sequences as in sequences/
│ │
│ └── detection_labels/ 				Object detection labels
│ ├── train/
│ │ └── Scene_A/
│ │ └── <sequence_name>/ 				Directory for each sequence, containing a detection labels file per image/sample
│ │ └── <sample_timestamp>.txt 				Detection ground truth; format specifies object class and bounding box coordinates per image
│ └── ... 						Repeat structure for additional scenes and sequences as in sequences/
|
├── event_samples/ 					Sampled event data files at different batch-sampling durations
│ └── <sample_duration>/
│ ├── train/
│ │ └── Scene_A/
│ │ └── <sequence_name>/ 				Contains event data files sampled according to the specified duration
│ │ └── <sample_timestamp>.aedat 			Sampled event data file encoded in .aedat format
│ └── ... 						Repeat structure for additional scenes and sequences as in sequences/
|
└── data_splits/ 					CSV files for indexing dataset splits for object detection
  ├── MEVDT_<sample_duration>_train.csv 		Training set index file for each sampling duration {original rate, 100, 200, 500 ms}
  └── MEVDT_<sample_duration>_test.csv 			Testing set index file for each sampling duration {original rate, 100, 200, 500 ms}