Kosmos-2.5
Microsoft Document AI | GitHub
Model description
Kosmos-2.5 is a multimodal literate model for machine reading of text-intensive images. Pre-trained on large-scale text-intensive images, Kosmos-2.5 excels in two distinct yet cooperative transcription tasks: (1) generating spatially-aware text blocks, where each block of text is assigned its spatial coordinates within the image, and (2) producing structured text output that captures styles and structures into the markdown format. This unified multimodal literate capability is achieved through a shared decoder-only auto-regressive Transformer architecture, task-specific prompts, and flexible text representations. We evaluate Kosmos-2.5 on end-to-end document-level text recognition and image-to-markdown text generation. Furthermore, the model can be readily adapted for any text-intensive image understanding task with different prompts through supervised fine-tuning, making it a general-purpose tool for real-world applications involving text-rich images. This work also paves the way for the future scaling of multimodal large language models.
Kosmos-2.5: A Multimodal Literate Model
NOTE:
Since this is a generative model, there is a risk of hallucination during the generation process, and it CAN NOT guarantee the accuracy of all OCR/Markdown results in the images.
Inference
KOSMOS-2.5 is supported from Transformers >= 4.56. Find the docs here.
Markdown Task: For usage instructions, please refer to md.py.
import re
import torch
import requests
from PIL import Image, ImageDraw
from transformers import AutoProcessor, Kosmos2_5ForConditionalGeneration, infer_device
repo = "microsoft/kosmos-2.5"
device = "cuda:0"
dtype = torch.bfloat16
model = Kosmos2_5ForConditionalGeneration.from_pretrained(repo, device_map=device, dtype=dtype)
processor = AutoProcessor.from_pretrained(repo)
# sample image
url = "https://huggingface.co/microsoft/kosmos-2.5/resolve/main/receipt_00008.png"
image = Image.open(requests.get(url, stream=True).raw)
prompt = "<md>"
inputs = processor(text=prompt, images=image, return_tensors="pt")
height, width = inputs.pop("height"), inputs.pop("width")
raw_width, raw_height = image.size
scale_height = raw_height / height
scale_width = raw_width / width
inputs = {k: v.to(device) if v is not None else None for k, v in inputs.items()}
inputs["flattened_patches"] = inputs["flattened_patches"].to(dtype)
generated_ids = model.generate(
**inputs,
max_new_tokens=1024,
)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)
print(generated_text[0])
OCR Task: For usage instructions, please refer to ocr.py.
import re
import torch
import requests
from PIL import Image, ImageDraw
from transformers import AutoProcessor, Kosmos2_5ForConditionalGeneration, infer_device
repo = "microsoft/kosmos-2.5"
device = "cuda:0"
dtype = torch.bfloat16
model = Kosmos2_5ForConditionalGeneration.from_pretrained(repo, device_map=device, dtype=dtype)
processor = AutoProcessor.from_pretrained(repo)
# sample image
url = "https://huggingface.co/microsoft/kosmos-2.5/resolve/main/receipt_00008.png"
image = Image.open(requests.get(url, stream=True).raw)
# bs = 1
prompt = "<ocr>"
inputs = processor(text=prompt, images=image, return_tensors="pt")
height, width = inputs.pop("height"), inputs.pop("width")
raw_width, raw_height = image.size
scale_height = raw_height / height
scale_width = raw_width / width
# bs > 1, batch generation
# inputs = processor(text=[prompt, prompt], images=[image,image], return_tensors="pt")
# height, width = inputs.pop("height"), inputs.pop("width")
# raw_width, raw_height = image.size
# scale_height = raw_height / height[0]
# scale_width = raw_width / width[0]
inputs = {k: v.to(device) if v is not None else None for k, v in inputs.items()}
inputs["flattened_patches"] = inputs["flattened_patches"].to(dtype)
generated_ids = model.generate(
**inputs,
max_new_tokens=1024,
)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)
def post_process(y, scale_height, scale_width):
y = y.replace(prompt, "")
if "<md>" in prompt:
return y
pattern = r"<bbox><x_\d+><y_\d+><x_\d+><y_\d+></bbox>"
bboxs_raw = re.findall(pattern, y)
lines = re.split(pattern, y)[1:]
bboxs = [re.findall(r"\d+", i) for i in bboxs_raw]
bboxs = [[int(j) for j in i] for i in bboxs]
info = ""
for i in range(len(lines)):
box = bboxs[i]
x0, y0, x1, y1 = box
if not (x0 >= x1 or y0 >= y1):
x0 = int(x0 * scale_width)
y0 = int(y0 * scale_height)
x1 = int(x1 * scale_width)
y1 = int(y1 * scale_height)
info += f"{x0},{y0},{x1},{y0},{x1},{y1},{x0},{y1},{lines[i]}"
return info
output_text = post_process(generated_text[0], scale_height, scale_width)
print(output_text)
draw = ImageDraw.Draw(image)
lines = output_text.split("\n")
for line in lines:
# draw the bounding box
line = list(line.split(","))
if len(line) < 8:
continue
line = list(map(int, line[:8]))
draw.polygon(line, outline="red")
image.save("output.png")
Citation
If you find Kosmos-2.5 useful in your research, please cite the following paper:
@article{lv2023kosmos,
title={Kosmos-2.5: A multimodal literate model},
author={Lv, Tengchao and Huang, Yupan and Chen, Jingye and Cui, Lei and Ma, Shuming and Chang, Yaoyao and Huang, Shaohan and Wang, Wenhui and Dong, Li and Luo, Weiyao and others},
journal={arXiv preprint arXiv:2309.11419},
year={2023}
}
License
The content of this project itself is licensed under the MIT