jddlmZddlZddlZddlmZddlmZmZddl m Z m Z m Z ddd Z e dz ddfddZedkr e ddSdS)) annotationsN)Path) IMG_FORMATSimg2label_paths) DATASETS_DIRLOGGERTQDM皙? source_dir str | Path train_ratiofloatreturnrc t|}t|d}|dz |dz }}|d|d|dd|D}td|D}t |d|d }t jd |d |d |d dd|z d d |D]} || jz d|| jz dt| d} tj | tt | |z} | d| D]'} tj| || jz | jz (| | dD]'} tj| || jz | jz (t jd|d|S)u)Split classification dataset into train and val directories in a new directory. Creates a new directory '{source_dir}_split' with train/val subdirectories, preserving the original class structure with an 80/20 split by default. Directory structure: Before: caltech/ ├── class1/ │ ├── img1.jpg │ ├── img2.jpg │ └── ... ├── class2/ │ ├── img1.jpg │ └── ... └── ... After: caltech_split/ ├── train/ │ ├── class1/ │ │ ├── img1.jpg │ │ └── ... │ ├── class2/ │ │ ├── img1.jpg │ │ └── ... │ └── ... └── val/ ├── class1/ │ ├── img2.jpg │ └── ... ├── class2/ │ └── ... └── ... Args: source_dir (str | Path): Path to classification dataset root directory. train_ratio (float): Ratio for train split, between 0 and 1. Returns: (Path): Path to the created split directory. Examples: Split dataset with default 80/20 ratio >>> split_classify_dataset("path/to/caltech") Split with custom ratio >>> split_classify_dataset("path/to/caltech", 0.75) _splittrainvalT)exist_okc:g|]}||S)is_dir.0ds [/home/longshao/multi-rider-rag/.venv/lib/python3.11/site-packages/ultralytics/data/split.py z*split_classify_dataset..Is%AAAahhjjA!AAAc3tK|]3}tt|dV4dS)*.*N)lenlistglobrs r z)split_classify_dataset..Js<DDAs4u ..//DDDDDDrz classes, z imagesz Splitting z (z) into z.0%z train, z val...rNzSplit complete in u ✅)rmkdiriterdirsumr rinfonamer!r"randomshuffleintshutilcopy2) r r source_path split_path train_pathval_path class_dirs total_imagesstats class_dir image_files split_idximgs rsplit_classify_datasetr: sPdz""K,,,--J%/e1CJd###d### NNDN!!!BA[0022AAAJDDDDDDDL: ? ?, ? ? ?E Ko[ooEoo+oooSTWbSboooopppDD in $++T+::: IN "))4)8889>>%0011 {###K((;677 z z* F FC Lj9>9CHD E E E Eyzz* D DC Lh7#(B C C C C D K5Z555666 rz coco8/images)g?g?gFpathweightstuple[float, float, float]annotated_onlyboolNonec|t|}td|dD}t|}t jdt jgd||}gd}|D]:}|j|z r|j|z ;tj d|d|zztt||| D]\}} |rBttt| gdr}t!|j||z d d 5} | d | |jdzdddn #1swxYwYdS)aAutomatically split a dataset into train/val/test splits and save the resulting splits into autosplit_*.txt files. Args: path (Path): Path to images directory. weights (tuple[float, float, float]): Train, validation, and test split fractions. annotated_only (bool): If True, only images with an associated txt file are used. Examples: Split images with default weights >>> from ultralytics.data.split import autosplit >>> autosplit() Split with custom weights and annotated images only >>> autosplit(path="path/to/images", weights=(0.8, 0.15, 0.05), annotated_only=True) c3jK|].}|jddtv*|V/dS)r$N)suffixlowerr)rxs rr#zautosplit..xsBWW18ABB<3E3E3G3G;3V3V13V3V3V3VWWrrr)rr$)r<k)zautosplit_train.txtzautosplit_val.txtzautosplit_test.txtzAutosplitting images from z!, using *.txt labeled images only)totalazutf-8)encodingz./ N)rsortedrglobr r*seedchoicesparentexistsunlinkrr(r ziprstropenwrite relative_toas_posix) r;r<r>filesnindicestxtrEir9fs r autosplitr_bs * ::D WWdjj//WWW W WE E A KNNNnYYY1===G L L LC '' K!O # # % % ' [1_ $ $ & & & K3T336Y\j6jjkkks7E**!444OO3 Oos3xxj&A&A!&D!E!E!L!L!N!N OdkCF*C'BBB OaFS__T[99BBDDFFMNNN O O O O O O O O O O O O O O OOOsAF00F4 7F4 __main__ caltech101)r )r r r rrr)r;rr<r=r>r?rr@) __future__rr*r-pathlibrultralytics.data.utilsrrultralytics.utilsrrr r:r___name__rrrrgs#""""" ????????8888888888RRRRRl.*9 $O$O$O$O$ON z<(((((r