Spaces:
Running
Running
File size: 3,133 Bytes
e6f24ae | 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 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 | """Abstract base class for style datasets."""
import logging
from abc import ABC, abstractmethod
from pathlib import Path
from typing import Any, Dict, List, Optional, Tuple
from PIL import Image
logger = logging.getLogger(__name__)
class StyleDataset(ABC):
"""Base class for all style captioning datasets.
Each dataset provides:
- Images for probing (train split) and evaluation (test split)
- Style labels
- Optionally: ground-truth styled captions (for GT metrics)
"""
def __init__(
self,
data_dir: str,
split: str = "test",
n_images: Optional[int] = None,
seed: int = 42,
):
self.data_dir = Path(data_dir)
self.split = split
self.n_images = n_images
self.seed = seed
self._data: Optional[List[Dict[str, Any]]] = None
@property
@abstractmethod
def track(self) -> str:
"""Track letter (A, B, C, D)."""
...
@property
@abstractmethod
def styles(self) -> List[str]:
"""List of style names in this track."""
...
@property
@abstractmethod
def has_ground_truth(self) -> bool:
"""Whether this dataset has ground-truth styled captions."""
...
@abstractmethod
def _load_data(self) -> List[Dict[str, Any]]:
"""Load raw data from disk.
Returns list of dicts with at minimum:
- "image_id": str or int
- "image_path": str (absolute path to image)
- "style": str (style name)
- "caption_gt": Optional[List[str]] (ground-truth captions, if available)
"""
...
@property
def data(self) -> List[Dict[str, Any]]:
"""Lazy-loaded data."""
if self._data is None:
self._data = self._load_data()
return self._data
def get_images(self, style: str) -> List[Dict[str, Any]]:
"""Get all items for a given style."""
items = [d for d in self.data if d["style"] == style]
if self.n_images is not None:
import random
rng = random.Random(self.seed)
items = rng.sample(items, min(self.n_images, len(items)))
return items
def load_image(self, image_path: str) -> Image.Image:
"""Load a PIL Image from path."""
return Image.open(image_path).convert("RGB")
def get_ground_truth(self, image_id: str, style: str) -> Optional[List[str]]:
"""Get ground-truth captions for an image+style pair."""
if not self.has_ground_truth:
return None
items = [d for d in self.data if d["image_id"] == image_id and d["style"] == style]
if not items:
return None
refs = []
for item in items:
if item.get("caption_gt"):
refs.extend(item["caption_gt"] if isinstance(item["caption_gt"], list) else [item["caption_gt"]])
return refs if refs else None
def __len__(self) -> int:
return len(self.data)
def __repr__(self) -> str:
return f"{self.__class__.__name__}(track={self.track}, split={self.split}, n={len(self)})"
|