timm/timm_image.py

# Copyright 2021 Zilliz. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import logging
import numpy

from towhee.operator.base import NNOperator, OperatorFlag
from towhee import register

import torch
from torch import nn

from PIL import Image as PILImage

from timm.data.transforms_factory import create_transform
from timm.data import resolve_data_config
from timm.models.factory import create_model

import warnings

warnings.filterwarnings('ignore')


log = logging.getLogger()

@register(output_schema=['vec'])
class TimmImage(NNOperator):
    """
    Pytorch image embedding operator that uses the Pytorch Image Model (timm) collection.
    Args:
        model_name (`str`):
            Which model to use for the embeddings.
        num_classes (`int = 1000`):
            Number of classes for classification.
    """

    def __init__(self, model_name: str, num_classes: int = 1000) -> None:
        super().__init__()
        self.device = 'cuda' if torch.cuda.is_available() else 'cpu'
        self.model = create_model(model_name, pretrained=True, num_classes=num_classes)
        self.model.to(self.device)
        self.model.eval()
        config = resolve_data_config({}, model=self.model)
        self.tfms = create_transform(**config)

    def __call__(self, img: numpy.ndarray) -> numpy.ndarray:
        if hasattr(img, 'mode'):
            if img.mode != 'RGB':
                log.error(f'Invalid image mode: expect "RGB" but receive "{img.mode}".')
                raise AssertionError(f'Invalid image mode "{img.mode}".')
        else:
            log.warning(f'Image mode is not specified. Using "RGB" now.')

        img = PILImage.fromarray(img.astype('uint8'), 'RGB')
        img = self.tfms(img).unsqueeze(0)
        img = img.to(self.device)
        features = self.model.forward_features(img)
        if features.dim() == 4:
            global_pool = nn.AdaptiveAvgPool2d(1)
            features = global_pool(features)

        features = features.to('cpu')
        feature_vector = features.flatten().detach().numpy()
        return feature_vector


# if __name__ == '__main__':
#     import cv2
#     from towhee._types import Image
#
#
#     path = '/path/to/image'
#     img = cv2.imread(path)
#     img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
#     img = Image(img, 'RGB')
#
#     op = TimmImage('resnet50')
#     out = op(img)