Danbooruclip

🚀 DanbooruCLIP

A fine - tuned CLIP (ViT - L/14) model using the danburoo2021 dataset for vision - related tasks

🚀 Quick Start

The following is a basic example of using the model:

from PIL import Image
import requests

from transformers import CLIPProcessor, CLIPModel

model = CLIPModel.from_pretrained("OysterQAQ/DanbooruCLIP")
processor = CLIPProcessor.from_pretrained("OysterQAQ/DanbooruCLIP")

url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)

inputs = processor(text=["a photo of a cat", "a photo of a dog"], images=image, return_tensors="pt", padding=True)

outputs = model(**inputs)
logits_per_image = outputs.logits_per_image # this is the image - text similarity score
probs = logits_per_image.softmax(dim=1) # we can take the softmax to get the label probabilities

✨ Features

• Updated on July 17, 2023: Added the pixiv dataset for training.
• Fine - tuned the CLIP (ViT - L/14) model using the danburoo2021 dataset.
• Different learning rates and weight decays were used for different epochs during training.

🔧 Technical Details

Training Parameters

• For epochs 0 - 3, the learning rate is 4e - 6, and the weight decay is 1e - 3.
• For epochs 4 - 8, the learning rate is 1e - 6, and the weight decay is 1e - 3.

Label Pre - processing

            for i in range(length):
                # Load and scale the image
                if not is_image(data_from_db.path[i]):
                    continue

                try:
                    img = self.preprocess(
                        Image.open(data_from_db.path[i].replace("./", "/mnt/lvm/danbooru2021/danbooru2021/")))
                except Exception as e:
                    #print(e)
                    continue
                # Process the labels
                tags = json.loads(data_from_db.tags[i])
                # Prioritize character and work labels
                category_group = {}
                for tag in tags:
                    category_group.setdefault(tag["category"], []).append(tag)

                # category_group=groupby(tags, key=lambda x: (x["category"]))
                character_list = category_group[4] if 4 in category_group else []
                # Works need to filter those starting with 'bad'

                work_list = list(filter(
                    lambda e:
                               e["name"] != "original"
                            , category_group[3])) if 3 in category_group else []
                # work_list=  category_group[5] if 5 in category_group else []
                general_list = category_group[0] if 0 in category_group else []
                caption = ""
                caption_2 = None
                for character in character_list:
                    if len(work_list) != 0:
                        # Remove the work content in parentheses
                        character["name"] = re.sub(u"\\(.*?\\)", "", character["name"])
                    caption += character["name"].replace("_", " ")
                    caption += ","
                caption = caption[:-1]
                caption += " "
                if len(work_list) != 0:
                    caption += "from "
                for work in work_list:
                    caption += work["name"].replace("_", " ")
                    caption += " "
                # General tags
                if len(general_list) != 0:
                    caption += "with "
                if len(general_list) > 20:
                    general_list_1 = general_list[:int(len(general_list) / 2)]
                    general_list_2 = general_list[int(len(general_list) / 2):]
                    caption_2 = caption
                    for general in general_list_1:
                        if general["name"].find("girl") == -1 and general["name"].find("boy") == -1 and len(
                                re.findall(is_contain, general["name"])) != 0:
                            caption_2 += general["name"].replace("_", " ")
                            caption_2 += ","
                    caption_2 = caption_2[:-1]
                    for general in general_list_2:
                        if general["name"].find("girl") == -1 and general["name"].find("boy") == -1 and len(
                                re.findall(is_contain, general["name"])) != 0:
                            caption += general["name"].replace("_", " ")
                            caption += ","
                    caption = caption[:-1]
                else:
                    for general in general_list:
                        # If the number of label data is greater than 20, split it into two captions
                        if general["name"].find("girl") == -1 and general["name"].find("boy") == -1 and len(
                                re.findall(is_contain, general["name"])) != 0:
                            caption += general["name"].replace("_", " ")
                            caption += ","
                    caption = caption[:-1]

                # Summarize the labels into sentences
                # Tokenize the sentences
                # Return
                # Truncate if it is too long, otherwise use Hugging Face's method
                text_1 = clip.tokenize(texts=caption, truncate=True)
                text_2= None
                if caption_2 is not None:
                    text_2 = clip.tokenize(texts=caption_2, truncate=True)
                # Processing logic

                # print(img)
                yield img, text_1[0]
                if text_2 is not None:
                    yield img, text_2[0]

📄 Feedback

Where to send questions or comments about the model

Please use this Google Form