README.md

metadata

tags:
  - image-to-text
  - image-captioning
license: apache-2.0
language:
  - zh
widget:
  - src: >-
      https://datasets-server.huggingface.co/assets/Maciel/e-commerce-sample-images/--/Maciel--e-commerce-sample-images/train/2/image/image.jpg
    example_title: 小耳钉
  - src: >-
      https://datasets-server.huggingface.co/assets/Maciel/e-commerce-sample-images/--/Maciel--e-commerce-sample-images/train/5/image/image.jpg
    example_title: 短裙
  - src: >-
      https://datasets-server.huggingface.co/assets/Maciel/e-commerce-sample-images/--/Maciel--e-commerce-sample-images/train/7/image/image.jpg
    example_title: 高跟鞋

功能介绍

该模型功能主要是对图片生成文字描述。模型结构使用Encoder-Decoder结构，其中Encoder端使用BEiT模型，Decoder使用GPT模型。

使用中文Muge数据集训练语料，训练5k步，最终验证集loss为0.3737，rouge1为20.419，rouge2为7.3553，rougeL为17.3753，rougeLsum为17.376。

Github项目地址

如何使用

from transformers import VisionEncoderDecoderModel, ViTFeatureExtractor, AutoTokenizer
from PIL import Image

pretrained = "Maciel/Muge-Image-Caption"
model = VisionEncoderDecoderModel.from_pretrained(pretrained)
feature_extractor = ViTFeatureExtractor.from_pretrained(pretrained)
tokenizer = AutoTokenizer.from_pretrained(pretrained)

image_path = "https://huggingface.co/Maciel/Muge-Image-Caption/blob/main/%E9%AB%98%E8%B7%9F%E9%9E%8B.jpg"
image = Image.open(image_path)
if image.mode != "RGB":
        image = image.convert("RGB")
pixel_values = feature_extractor(images=image, return_tensors="pt").pixel_values

output_ids = model.generate(pixel_values, **gen_kwargs)
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
preds = [pred.strip() for pred in preds]
print(preds)