File size: 3,062 Bytes
61f3f56
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
from PIL import Image
from torchvision import transforms
from transformers import ProcessorMixin, BatchEncoding
from transformers.image_processing_utils import BatchFeature

OPENAI_DATASET_MEAN = (0.48145466, 0.4578275, 0.40821073)
OPENAI_DATASET_STD = (0.26862954, 0.26130258, 0.27577711)

def make_list_of_images(x):
    if not isinstance(x, list):
        return [x]
    return x

def get_thermal_transform(config):
    config = config.vision_config
    transform = transforms.Compose(
        [
            transforms.ToTensor(),
            transforms.Resize(224, interpolation=transforms.InterpolationMode.BICUBIC),
            transforms.CenterCrop(224),
            transforms.Normalize(OPENAI_DATASET_MEAN, OPENAI_DATASET_STD)  # assume image
        ]
    )
    return transform


def load_and_transform_thermal(thermal_path, transform):
    thermal = Image.open(thermal_path)
    thermal_outputs = transform(thermal)
    return thermal_outputs

class LanguageBindThermalProcessor(ProcessorMixin):
    attributes = []
    tokenizer_class = ("LanguageBindThermalTokenizer")

    def __init__(self, config, tokenizer=None, **kwargs):
        super().__init__(**kwargs)
        self.config = config
        self.transform = get_thermal_transform(config)
        self.image_processor = load_and_transform_thermal
        self.tokenizer = tokenizer

    def __call__(self, images=None, text=None, context_length=77, return_tensors=None, **kwargs):
        if text is None and images is None:
            raise ValueError("You have to specify either text or images. Both cannot be none.")

        if text is not None:
            encoding = self.tokenizer(text, max_length=context_length, padding='max_length',
                                      truncation=True, return_tensors=return_tensors, **kwargs)

        if images is not None:
            images = make_list_of_images(images)
            image_features = [self.image_processor(image, self.transform) for image in images]
            image_features = torch.stack(image_features)

        if text is not None and images is not None:
            encoding["pixel_values"] = image_features
            return encoding
        elif text is not None:
            return encoding
        else:
            return {"pixel_values": image_features}

    def batch_decode(self, skip_special_tokens=True, *args, **kwargs):
        """
        This method forwards all its arguments to CLIPTokenizerFast's [`~PreTrainedTokenizer.batch_decode`]. Please
        refer to the docstring of this method for more information.
        """
        return self.tokenizer.batch_decode(*args, skip_special_tokens=skip_special_tokens, **kwargs)

    def decode(self, skip_special_tokens=True, *args, **kwargs):
        """
        This method forwards all its arguments to CLIPTokenizerFast's [`~PreTrainedTokenizer.decode`]. Please refer to
        the docstring of this method for more information.
        """
        return self.tokenizer.decode(*args, skip_special_tokens=skip_special_tokens, **kwargs)