79 lines
3.4 KiB
Python
79 lines
3.4 KiB
Python
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
|
|
|
|
import cv2
|
|
import torch
|
|
from PIL import Image
|
|
|
|
from ultralytics.engine.predictor import BasePredictor
|
|
from ultralytics.engine.results import Results
|
|
from ultralytics.utils import DEFAULT_CFG, ops
|
|
|
|
|
|
class ClassificationPredictor(BasePredictor):
|
|
"""
|
|
A class extending the BasePredictor class for prediction based on a classification model.
|
|
|
|
This predictor handles the specific requirements of classification models, including preprocessing images
|
|
and postprocessing predictions to generate classification results.
|
|
|
|
Attributes:
|
|
args (Dict): Configuration arguments for the predictor.
|
|
_legacy_transform_name (str): Name of the legacy transform class for backward compatibility.
|
|
|
|
Methods:
|
|
preprocess: Convert input images to model-compatible format.
|
|
postprocess: Process model predictions into Results objects.
|
|
|
|
Notes:
|
|
- Torchvision classification models can also be passed to the 'model' argument, i.e. model='resnet18'.
|
|
|
|
Examples:
|
|
>>> from ultralytics.utils import ASSETS
|
|
>>> from ultralytics.models.yolo.classify import ClassificationPredictor
|
|
>>> args = dict(model="yolo11n-cls.pt", source=ASSETS)
|
|
>>> predictor = ClassificationPredictor(overrides=args)
|
|
>>> predictor.predict_cli()
|
|
"""
|
|
|
|
def __init__(self, cfg=DEFAULT_CFG, overrides=None, _callbacks=None):
|
|
"""Initialize the ClassificationPredictor with the specified configuration and set task to 'classify'."""
|
|
super().__init__(cfg, overrides, _callbacks)
|
|
self.args.task = "classify"
|
|
self._legacy_transform_name = "ultralytics.yolo.data.augment.ToTensor"
|
|
|
|
def preprocess(self, img):
|
|
"""Convert input images to model-compatible tensor format with appropriate normalization."""
|
|
if not isinstance(img, torch.Tensor):
|
|
is_legacy_transform = any(
|
|
self._legacy_transform_name in str(transform) for transform in self.transforms.transforms
|
|
)
|
|
if is_legacy_transform: # to handle legacy transforms
|
|
img = torch.stack([self.transforms(im) for im in img], dim=0)
|
|
else:
|
|
img = torch.stack(
|
|
[self.transforms(Image.fromarray(cv2.cvtColor(im, cv2.COLOR_BGR2RGB))) for im in img], dim=0
|
|
)
|
|
img = (img if isinstance(img, torch.Tensor) else torch.from_numpy(img)).to(self.model.device)
|
|
return img.half() if self.model.fp16 else img.float() # uint8 to fp16/32
|
|
|
|
def postprocess(self, preds, img, orig_imgs):
|
|
"""
|
|
Process predictions to return Results objects with classification probabilities.
|
|
|
|
Args:
|
|
preds (torch.Tensor): Raw predictions from the model.
|
|
img (torch.Tensor): Input images after preprocessing.
|
|
orig_imgs (List[np.ndarray] | torch.Tensor): Original images before preprocessing.
|
|
|
|
Returns:
|
|
(List[Results]): List of Results objects containing classification results for each image.
|
|
"""
|
|
if not isinstance(orig_imgs, list): # input images are a torch.Tensor, not a list
|
|
orig_imgs = ops.convert_torch2numpy_batch(orig_imgs)
|
|
|
|
preds = preds[0] if isinstance(preds, (list, tuple)) else preds
|
|
return [
|
|
Results(orig_img, path=img_path, names=self.model.names, probs=pred)
|
|
for pred, orig_img, img_path in zip(preds, orig_imgs, self.batch[0])
|
|
]
|