模型:

facebook/mask2former-swin-large-ade-panoptic

英文

Mask2Former

Mask2Former模型基于ADE20k全景分割数据集进行训练(大尺寸版本,采用Swin骨干网络)。该模型在论文 Masked-attention Mask Transformer for Universal Image Segmentation 中进行了介绍,并于 this repository 首次发布。

免责声明:发布Mask2Former模型的团队没有为该模型编写模型卡片,因此这份模型卡片是由Hugging Face团队编写的。

模型描述

Mask2Former通过预测一组掩码和相应的标签来处理实例分割、语义分割和全景分割这三个任务。因此,所有3个任务都被视为实例分割。Mask2Former通过(i)用更先进的多尺度可变形注意力Transformer替换像素解码器,(ii)采用具有掩码注意力的Transformer解码器,在不引入额外计算量的情况下提高性能,以及(iii)通过计算子采样点上的损失而不是整个掩码来提高训练效率,从而优于以前的SOTA模型 MaskFormer

应用与限制

您可以将此特定检查点用于全景分割。请查看 model hub 以寻找您感兴趣的其他任务的微调版本。

如何使用

以下是如何使用该模型的方法:

import requests
import torch
from PIL import Image
from transformers import AutoImageProcessor, Mask2FormerForUniversalSegmentation


# load Mask2Former fine-tuned on ADE20k panoptic segmentation
processor = AutoImageProcessor.from_pretrained("facebook/mask2former-swin-large-ade-panoptic")
model = Mask2FormerForUniversalSegmentation.from_pretrained("facebook/mask2former-swin-large-ade-panoptic")

url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)
inputs = processor(images=image, return_tensors="pt")

with torch.no_grad():
    outputs = model(**inputs)

# model predicts class_queries_logits of shape `(batch_size, num_queries)`
# and masks_queries_logits of shape `(batch_size, num_queries, height, width)`
class_queries_logits = outputs.class_queries_logits
masks_queries_logits = outputs.masks_queries_logits

# you can pass them to processor for postprocessing
result = processor.post_process_panoptic_segmentation(outputs, target_sizes=[image.size[::-1]])[0]
# we refer to the demo notebooks for visualization (see "Resources" section in the Mask2Former docs)
predicted_panoptic_map = result["segmentation"]

有关更多代码示例,请参阅 documentation