OneFormer

OneFormer模型基于COCO数据集进行训练（大型版本，Swin骨干）。该模型在Jain等人的论文 OneFormer: One Transformer to Rule Universal Image Segmentation 中首次提出，并于 this repository 首次发布。

模型描述

OneFormer是第一个多任务通用图像分割框架。它只需通过单一通用架构、单一模型和单一数据集进行训练，即可在语义分割、实例分割和全景分割任务上优于现有的专门化模型。OneFormer使用任务令牌来将模型与当前任务相关联，使得架构在训练时具有任务指导性，在推断时具有任务动态性，而所有这些都由单一模型完成。

预期用途和限制

您可以使用此特定检查点进行语义分割、实例分割和全景分割。请参阅 model hub 以查找在其他数据集上进行微调的其他版本。

如何使用

以下是如何使用此模型的方法：

from transformers import OneFormerProcessor, OneFormerForUniversalSegmentation
from PIL import Image
import requests
url = "https://huggingface.co/datasets/shi-labs/oneformer_demo/blob/main/coco.jpeg"
image = Image.open(requests.get(url, stream=True).raw)

# Loading a single model for all three tasks
processor = OneFormerProcessor.from_pretrained("shi-labs/oneformer_coco_swin_large")
model = OneFormerForUniversalSegmentation.from_pretrained("shi-labs/oneformer_coco_swin_large")

# Semantic Segmentation
semantic_inputs = processor(images=image, task_inputs=["semantic"], return_tensors="pt")
semantic_outputs = model(**semantic_inputs)
# pass through image_processor for postprocessing
predicted_semantic_map = processor.post_process_semantic_segmentation(outputs, target_sizes=[image.size[::-1]])[0]

# Instance Segmentation
instance_inputs = processor(images=image, task_inputs=["instance"], return_tensors="pt")
instance_outputs = model(**instance_inputs)
# pass through image_processor for postprocessing
predicted_instance_map = processor.post_process_instance_segmentation(outputs, target_sizes=[image.size[::-1]])[0]["segmentation"]

# Panoptic Segmentation
panoptic_inputs = processor(images=image, task_inputs=["panoptic"], return_tensors="pt")
panoptic_outputs = model(**panoptic_inputs)
# pass through image_processor for postprocessing
predicted_semantic_map = processor.post_process_panoptic_segmentation(outputs, target_sizes=[image.size[::-1]])[0]["segmentation"]

更多示例，请参阅 documentation 。

引用

@article{jain2022oneformer,
      title={{OneFormer: One Transformer to Rule Universal Image Segmentation}},
      author={Jitesh Jain and Jiachen Li and MangTik Chiu and Ali Hassani and Nikita Orlov and Humphrey Shi},
      journal={arXiv}, 
      year={2022}
    }