Oneformer Ade20k Dinat Large

🚀 OneFormer

OneFormer是首個多任務通用圖像分割框架，僅需在單個數據集上以單一通用架構和模型進行一次訓練，就能在語義、實例和全景分割任務中超越現有的專門模型。

🚀 快速開始

OneFormer模型在ADE20k數據集（大尺寸版本，Dinat主幹網絡）上進行訓練。它由Jain等人在論文 OneFormer: One Transformer to Rule Universal Image Segmentation 中提出，並首次在 此倉庫 中發佈。

✨ 主要特性

模型描述

OneFormer是首個多任務通用圖像分割框架。它僅需使用單一通用架構、單一模型，並在單一數據集上進行一次訓練，就能在語義、實例和全景分割任務中超越現有的專門模型。OneFormer使用任務令牌來使模型專注於當前任務，使架構在訓練時具有任務導向性，在推理時具有任務動態性，且所有操作都通過單一模型完成。

預期用途與侷限性

你可以使用此特定檢查點進行語義、實例和全景分割。請查看 模型中心 以查找在不同數據集上微調的其他版本。

💻 使用示例

基礎用法

from transformers import OneFormerProcessor, OneFormerForUniversalSegmentation
from PIL import Image
import requests
url = "https://huggingface.co/datasets/shi-labs/oneformer_demo/blob/main/ade20k.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_ade20k_dinat_large")
model = OneFormerForUniversalSegmentation.from_pretrained("shi-labs/oneformer_ade20k_dinat_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"]

如需更多示例，請參考 文檔。

📚 詳細文檔

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}
    }

📄 許可證

本項目採用MIT許可證。