Geometric Shapes Classification

🚀 几何形状分类模型

本项目的几何形状分类模型是一个图像分类的视觉语言编码器模型，它基于 google/siglip2-base-patch16-224 进行微调，用于多类形状识别任务。该模型使用 SiglipForImageClassification 架构对各种几何形状进行分类。

🚀 快速开始

安装依赖

!pip install -q transformers torch pillow gradio

运行代码

import gradio as gr
from transformers import AutoImageProcessor
from transformers import SiglipForImageClassification
from PIL import Image
import torch

# Load model and processor
model_name = "prithivMLmods/Geometric-Shapes-Classification"
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)

# Label mapping with symbols
labels = {
    "0": "Circle ◯",
    "1": "Kite ⬰",
    "2": "Parallelogram ▰",
    "3": "Rectangle ▭",
    "4": "Rhombus ◆",
    "5": "Square ◼",
    "6": "Trapezoid ⏢",
    "7": "Triangle ▲"
}

def classify_shape(image):
    """Classifies the geometric shape in the input image."""
    image = Image.fromarray(image).convert("RGB")
    inputs = processor(images=image, return_tensors="pt")

    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
        probs = torch.nn.functional.softmax(logits, dim=1).squeeze().tolist()

    predictions = {labels[str(i)]: round(probs[i], 3) for i in range(len(probs))}
    
    return predictions

# Gradio interface
iface = gr.Interface(
    fn=classify_shape,
    inputs=gr.Image(type="numpy"),
    outputs=gr.Label(label="Prediction Scores"),
    title="Geometric Shapes Classification",
    description="Upload an image to classify geometric shapes such as circle, triangle, square, and more."
)

# Launch the app
if __name__ == "__main__":
    iface.launch()

💻 使用示例

基础用法

# 以下代码展示了如何使用该模型进行几何形状分类
import gradio as gr
from transformers import AutoImageProcessor
from transformers import SiglipForImageClassification
from PIL import Image
import torch

# 加载模型和处理器
model_name = "prithivMLmods/Geometric-Shapes-Classification"
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)

# 带有符号的标签映射
labels = {
    "0": "Circle ◯",
    "1": "Kite ⬰",
    "2": "Parallelogram ▰",
    "3": "Rectangle ▭",
    "4": "Rhombus ◆",
    "5": "Square ◼",
    "6": "Trapezoid ⏢",
    "7": "Triangle ▲"
}

def classify_shape(image):
    """对输入图像中的几何形状进行分类。"""
    image = Image.fromarray(image).convert("RGB")
    inputs = processor(images=image, return_tensors="pt")

    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
        probs = torch.nn.functional.softmax(logits, dim=1).squeeze().tolist()

    predictions = {labels[str(i)]: round(probs[i], 3) for i in range(len(probs))}
    
    return predictions

# Gradio 界面
iface = gr.Interface(
    fn=classify_shape,
    inputs=gr.Image(type="numpy"),
    outputs=gr.Label(label="Prediction Scores"),
    title="Geometric Shapes Classification",
    description="上传一张图像，对圆形、三角形、正方形等几何形状进行分类。"
)

# 启动应用
if __name__ == "__main__":
    iface.launch()

📚 详细文档

分类报告

Classification Report:
                 precision    recall  f1-score   support

       Circle ◯     0.9921    0.9987    0.9953      1500
         Kite ⬰     0.9927    0.9927    0.9927      1500
Parallelogram ▰     0.9926    0.9840    0.9883      1500
    Rectangle ▭     0.9993    0.9913    0.9953      1500
      Rhombus ◆     0.9846    0.9820    0.9833      1500
       Square ◼     0.9914    0.9987    0.9950      1500
    Trapezoid ⏢     0.9966    0.9793    0.9879      1500
     Triangle ▲     0.9772    0.9993    0.9881      1500

       accuracy                         0.9908     12000
      macro avg     0.9908    0.9908    0.9907     12000
   weighted avg     0.9908    0.9908    0.9907     12000

模型分类的类别

该模型将图像分类为以下类别：

• 类别 0：圆形 ◯
• 类别 1：风筝形 ⬰
• 类别 2：平行四边形 ▰
• 类别 3：矩形 ▭
• 类别 4：菱形 ◆
• 类别 5：正方形 ◼
• 类别 6：梯形 ⏢
• 类别 7：三角形 ▲

预期用途

几何形状分类 模型旨在识别图像中的基本几何形状。示例用例如下：

• 教育工具：用于以可视化方式学习和教授几何知识。
• 计算机视觉项目：作为机器人或自动化中的形状检测器。
• 图像分析：识别图表或工程图纸中的符号。
• 辅助技术：支持视障应用中的形状识别。

📄 许可证

本项目采用 Apache-2.0 许可证。

📦 模型信息