"""
Overview:
Detect upper-half body in anime images.
Trained on dataset `deepghs/anime_halfbody_detection <https://huggingface.co/datasets/deepghs/anime_halfbody_detection>`_ with YOLOv8.
.. image:: halfbody_detect_demo.plot.py.svg
:align: center
This is an overall benchmark of all the halfbody detect models:
.. image:: halfbody_detect_benchmark.plot.py.svg
:align: center
.. note::
Please note that the primary purpose of this tool is to crop upper-body images from illustrations.
Therefore, the training data used mostly consists of single-person images, and **the performance
on images with multiple people is not guaranteed**. If you indeed need to process
images with multiple people, the recommended approach is to first use
the :func:`imgutils.detect.person.detect_person` function to crop individuals,
and then use this tool to obtain upper-body images.
"""
from functools import lru_cache
from typing import List, Tuple
from huggingface_hub import hf_hub_download
from ._yolo import _image_preprocess, _data_postprocess
from ..data import ImageTyping, load_image, rgb_encode
from ..utils import open_onnx_model
@lru_cache()
def _open_halfbody_detect_model(level: str = 's', version: str = 'v1.0'):
return open_onnx_model(hf_hub_download(
f'deepghs/anime_halfbody_detection',
f'halfbody_detect_{version}_{level}/model.onnx'
))
_LABELS = ["halfbody"]
[docs]def detect_halfbody(image: ImageTyping, level: str = 's', version: str = 'v1.0', max_infer_size=640,
conf_threshold: float = 0.5, iou_threshold: float = 0.7) \
-> List[Tuple[Tuple[int, int, int, int], str, float]]:
"""
Overview:
Detect human upper-half body in anime images.
:param image: Image to detect.
:param level: The model level being used can be either `s` or `n`.
The `n` model runs faster with smaller system overhead, while the `s` model achieves higher accuracy.
The default value is `s`.
:param version: Version of model, default is ``v1.0``.
:param max_infer_size: The maximum image size used for model inference, if the image size exceeds this limit,
the image will be resized and used for inference. The default value is `640` pixels.
:param conf_threshold: The confidence threshold, only detection results with confidence scores above
this threshold will be returned. The default value is `0.5`.
:param iou_threshold: The detection area coverage overlap threshold, areas with overlaps above this threshold
will be discarded. The default value is `0.7`.
:return: The detection results list, each item includes the detected area `(x0, y0, x1, y1)`,
the target type (always `halfbody`) and the target confidence score.
Examples::
>>> from imgutils.detect import detect_halfbody, detection_visualize
>>>
>>> image = 'squat.jpg'
>>> result = detect_halfbody(image) # detect it
>>> result
[((127, 21, 629, 637), 'halfbody', 0.9040350914001465)]
>>>
>>> # visualize it
>>> from matplotlib import pyplot as plt
>>> plt.imshow(detection_visualize(image, result))
>>> plt.show()
"""
image = load_image(image, mode='RGB')
new_image, old_size, new_size = _image_preprocess(image, max_infer_size)
data = rgb_encode(new_image)[None, ...]
output, = _open_halfbody_detect_model(level, version).run(['output0'], {'images': data})
return _data_postprocess(output[0], conf_threshold, iou_threshold, old_size, new_size, _LABELS)