Datasets:
Tasks:
Image Segmentation
Formats:
json
Sub-tasks:
semantic-segmentation
Languages:
English
Size:
1K - 10K
License:
File size: 1,508 Bytes
ed72660 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 | # AeroScapes — Ultralytics YOLO semantic segmentation dataset config
#
# This file lives INSIDE the dataset root itself (next to images/ and masks/)
# so the whole folder is self-contained and portable — no absolute machine
# path to edit. There is intentionally NO 'path:' key: when omitted,
# Ultralytics resolves train/val relative to this file's own directory
# (rather than joining a relative 'path:' against the global DATASETS_DIR
# setting), so this works correctly wherever the folder is placed or
# downloaded to.
#
# Format: https://docs.ultralytics.com/datasets/semantic/
# Masks: single-channel PNG, pixel value = class index (0-11). 255 is
# reserved for genuinely unrecognized values (none expected here —
# see the source repo's convert_aeroscapes_to_yolo.py, which
# validates every mask pixel at conversion time).
train: images/train # 2621 images
val: images/val # 648 images
# No 'test:' key — the source AeroScapes distribution has no test split
# (only ImageSets/trn.txt and val.txt).
masks_dir: masks # mirrors images/ structure: images/train -> masks/train, etc.
nc: 12
names:
0: Background
1: Person
2: Bike
3: Car
4: Drone
5: Boat
6: Animal
7: Obstacle
8: Construction
9: Vegetation
10: Road
11: Sky
# Example training command (run with this file's directory as --dst, or
# point data= at this file's absolute path from anywhere):
# yolo semantic train model=yolo26n-sem.pt data=data.yaml imgsz=896 epochs=120 batch=8
|