| ---
|
| language:
|
| - zh
|
| license: mit
|
| tags:
|
| - computer-vision
|
| - image-classification
|
| - convnextv2
|
| - insect-identification
|
| - agriculture
|
| - pest-detection
|
| datasets:
|
| - custom
|
| widget:
|
| - src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/tiger.jpg
|
| example_title: Tiger
|
| - src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/teapot.jpg
|
| example_title: Teapot
|
| - src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/palace.jpg
|
| example_title: Palace
|
| ---
|
|
|
| # ConvNextV2 玉米害虫识别模型
|
|
|
| 这是一个基于ConvNextV2架构训练的玉米害虫图像分类模型，专门用于识别8种常见玉米害虫。
|
|
|
| ## 模型详情
|
|
|
| - **模型架构**: ConvNextV2 Tiny
|
| - **输入尺寸**: 224x224 RGB图像
|
| - **输出类别**: 8种玉米害虫
|
| - **训练数据**: 自定义玉米害虫数据集
|
| - **用途**: 农业害虫监测与识别
|
|
|
| ## 类别标签
|
|
|
| [
|
| "草地贪夜蛾",
|
| "赤须盲蝽",
|
| "点蜂缘蝽",
|
| "龟纹瓢虫",
|
| "棉铃虫",
|
| "茄二十八星瓢虫",
|
| "双斑长跗萤叶甲",
|
| "玉米螟"
|
| ]
|
|
|
| ## 使用方法
|
|
|
| ### 使用Transformers库
|
|
|
| ```python
|
| from transformers import AutoImageProcessor, AutoModelForImageClassification
|
| import torch
|
| from PIL import Image
|
|
|
| # 加载模型和处理器
|
| model_name = "Jaysaly/cornlarva_convnextv2"
|
| processor = AutoImageProcessor.from_pretrained(model_name)
|
| model = AutoModelForImageClassification.from_pretrained(model_name)
|
|
|
| # 准备图像
|
| image = Image.open("path_to_image.jpg").convert("RGB")
|
| inputs = processor(images=image, return_tensors="pt")
|
|
|
| # 推理
|
| with torch.no_grad():
|
| outputs = model(**inputs)
|
| logits = outputs.logits
|
| predicted_class_idx = logits.argmax(-1).item()
|
|
|
| # 获取预测结果
|
| predicted_label = model.config.id2label[str(predicted_class_idx)]
|
| print(f"预测结果: {predicted_label}")
|
| ```
|
|
|
| ### 使用Hugging Face Inference API
|
|
|
| ```python
|
| import requests
|
|
|
| API_URL = "https://api-inference.huggingface.co/models/Jaysaly/cornlarva_convnextv2"
|
| headers = {"Authorization": "Bearer YOUR_HF_TOKEN"}
|
|
|
| def query(image_bytes):
|
| response = requests.post(API_URL, headers=headers, data=image_bytes)
|
| return response.json()
|
| ```
|
|
|
| ## 训练信息
|
|
|
| - **框架**: PyTorch
|
| - **优化器**: AdamW
|
| - **学习率**: 1e-4
|
| - **批次大小**: 32
|
| - **训练周期**: 100
|
| - **数据增强**: 随机裁剪、水平翻转、颜色抖动
|
|
|
| ## 应用场景
|
|
|
| 1. **智能农业监测**: 自动识别田间害虫
|
| 2. **病虫害预警**: 及时发现害虫爆发
|
| 3. **精准施药**: 针对特定害虫采取防治措施
|
| 4. **科研教育**: 害虫识别教学与研究
|
|
|
| ## 性能指标
|
|
|
| 在验证集上的Top-1准确率: ~95%
|
|
|
| ## 限制
|
|
|
| - 仅支持训练数据中的8种玉米害虫
|
| - 需要清晰的昆虫特写图像
|
| - 对背景复杂的图片识别效果可能下降
|
|
|
| ## 作者
|
|
|
| Jaysaly
|
|
|
| ## 许可证
|
|
|
| MIT License
|
|
|
