模型描述 (Model Description)

?Paper | ?️Video |

运行环境 (Operating environment)

conda activate <your_env>
git clone https://github.com/IDEA-Research/GroundingDINO.git
cd GroundingDINO
pip install -e .

代码范例 (Code example)

from modelscope.pipelines import pipeline
import torch
import os
from PIL import Image, ImageDraw, ImageFont
import numpy as np
import requests
from io import BytesIO

def plot_boxes_to_image(image_pil, tgt):
    H, W = tgt["size"]
    boxes = tgt["boxes"]
    labels = tgt["labels"]
    assert len(boxes) == len(labels), "boxes and labels must have same length"

    draw = ImageDraw.Draw(image_pil)
    mask = Image.new("L", image_pil.size, 0)
    mask_draw = ImageDraw.Draw(mask)

    # draw boxes and masks
    for box, label in zip(boxes, labels):
        # from 0..1 to 0..W, 0..H
        box = box * torch.Tensor([W, H, W, H])
        # from xywh to xyxy
        box[:2] -= box[2:] / 2
        box[2:] += box[:2]
        # random color
        color = tuple(np.random.randint(0, 255, size=3).tolist())
        # draw
        x0, y0, x1, y1 = box
        x0, y0, x1, y1 = int(x0), int(y0), int(x1), int(y1)

        draw.rectangle([x0, y0, x1, y1], outline=color, width=6)
        # draw.text((x0, y0), str(label), fill=color)

        font = ImageFont.load_default()
        if hasattr(font, "getbbox"):
            bbox = draw.textbbox((x0, y0), str(label), font)
        else:
            w, h = draw.textsize(str(label), font)
            bbox = (x0, y0, w + x0, y0 + h)
        # bbox = draw.textbbox((x0, y0), str(label))
        draw.rectangle(bbox, fill=color)
        draw.text((x0, y0), str(label), fill="white")

        mask_draw.rectangle([x0, y0, x1, y1], fill=255, width=6)

    return image_pil, mask

pipe = pipeline('tiefang-grounding-dino-task',
                model='goldsj/GroundingDINO', model_revision='v1.0.0')


input_image_url = "https://ofasys.oss-cn-zhangjiakou.aliyuncs.com/data/coco/2014/val2014/COCO_val2014_000000222628.jpg"
response = requests.get(input_image_url)
input_image_file =BytesIO(response.content)

data = {
    'img': input_image_file,
    'caption': "person"
}
boxes_filt, pred_phrases = pipe(data)

# visualize pred

image_pil = Image.open(input_image_file).convert("RGB")

size = image_pil.size
pred_dict = {
    "boxes": boxes_filt,
    "size": [size[1], size[0]],  # H,W
    "labels": pred_phrases,
}
image_with_box = plot_boxes_to_image(image_pil, pred_dict)[0]
image_with_box.save("./pred.jpg")

Citation

If you find our work helpful for your research, please consider citing the following BibTeX entry.

@inproceedings{ShilongLiu2023GroundingDM,
  title={Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection},
  author={Shilong Liu and Zhaoyang Zeng and Tianhe Ren and Feng Li and Hao Zhang and Jie Yang and Chunyuan Li and Jianwei Yang and Hang Su and Jun Zhu and Lei Zhang},
  year={2023}
}