feat: 初始化香蕉叶病害智能识别系统

基于 ViT 视觉变换器实现香蕉叶片病害分类应用,支持 7 类病害识别。

  - app.py: Streamlit 前端应用,包含图片上传、病害识别、置信度图表展示及防治建议
  - pyproject.toml: 项目配置,声明 Python 3.14 及依赖(streamlit、transformers、torch、plotly 等)
  - uv.lock: 依赖版本锁定文件
  - Dockerfile: 基于 uv 镜像的容器化部署配置,含健康检查
  - justfile: 常用开发任务脚本(run、install、format、check 等)
  - main.py: 命令行入口
  - .doc/: 模型使用说明文档

  模型来源: Dmitry43243242/banana-disease-leaf-model (HuggingFace)
This commit is contained in:
zhenghu
2026-04-15 11:25:57 +08:00
parent f939f415a9
commit 86541eb55e
7 changed files with 1689 additions and 0 deletions

284
app.py Normal file
View File

@@ -0,0 +1,284 @@
"""
香蕉叶病害智能识别系统
基于 ViT 视觉变换器的香蕉叶片病害分类应用
"""
import os
import streamlit as st
import plotly.graph_objects as go
from PIL import Image
# 国内 HuggingFace 镜像加速
os.environ.setdefault("HF_ENDPOINT", "https://hf-mirror.com")
from transformers import pipeline # noqa: E402
# ─── Disease Label Mapping ──────────────────────────────────────────────────
# 模型输出 LABEL_0 ~ LABEL_6映射为实际病害名称
DISEASE_INFO = {
"LABEL_0": {
"name": "细菌性枯萎病",
"name_en": "Banana Bacterial Wilt (BBW)",
"color": "#e74c3c",
"description": "由细菌引起,叶片从边缘开始变黄枯萎,最终整株死亡。",
"advice": "及时清除病株,避免刀具交叉感染,使用无菌工具操作,加强田间排水。",
},
"LABEL_1": {
"name": "黑条斑病",
"name_en": "Black Sigatoka",
"color": "#8e44ad",
"description": "真菌性病害,叶片出现黑色条纹,严重时叶片大面积枯死。",
"advice": "选用抗病品种,合理密植保持通风,及时清除病叶,适时喷施杀菌剂。",
},
"LABEL_2": {
"name": "苞片花叶病毒病",
"name_en": "Bract Mosaic Virus (BBMV)",
"color": "#e67e22",
"description": "病毒性病害,苞片和叶片出现花叶状斑纹,影响果实品质。",
"advice": "使用无病种苗,防治传毒蚜虫,及时拔除病株,避免机械传播。",
},
"LABEL_3": {
"name": "健康叶片",
"name_en": "Healthy",
"color": "#27ae60",
"description": "叶片状态良好,无明显病害症状,颜色鲜绿。",
"advice": "继续保持良好的田间管理,合理施肥灌溉,定期巡检。",
},
"LABEL_4": {
"name": "花叶病毒病",
"name_en": "Banana Mosaic Virus",
"color": "#f39c12",
"description": "病毒性病害,叶片出现黄绿相间的花叶症状,植株矮化。",
"advice": "选用无毒组培苗,控制蚜虫等传毒媒介,拔除病株,工具消毒。",
},
"LABEL_5": {
"name": "巴拿马病(枯萎病)",
"name_en": "Panama Disease (Fusarium Wilt)",
"color": "#c0392b",
"description": "由尖孢镰刀菌引起,维管束变褐,叶片从下部开始黄化枯萎。",
"advice": "种植抗病品种,严格检疫,病区轮作或休耕,避免土壤传播。",
},
"LABEL_6": {
"name": "黄条斑病",
"name_en": "Yellow Sigatoka",
"color": "#d4ac0d",
"description": "真菌性病害,叶片出现黄色条纹,比黑条斑病症状较轻。",
"advice": "保持田间通风透光,清除病残叶,适时用药防治,合理施肥增强抗性。",
},
}
# ─── Page Config ────────────────────────────────────────────────────────────
st.set_page_config(
page_title="香蕉叶病害识别",
page_icon="🍌",
layout="wide",
initial_sidebar_state="expanded",
)
st.markdown("""
<style>
html, body, [class*="css"] {
font-family: "PingFang SC", "Microsoft YaHei", "Noto Sans SC", sans-serif;
}
</style>
""", unsafe_allow_html=True)
# ─── Model Loading ──────────────────────────────────────────────────────────
@st.cache_resource
def load_model():
"""加载 HuggingFace 模型(首次运行自动下载,约 343MB"""
classifier = pipeline(
"image-classification",
model="Dmitry43243242/banana-disease-leaf-model",
)
return classifier
# ─── Sidebar ────────────────────────────────────────────────────────────────
with st.sidebar:
st.header("🍌 香蕉叶病害识别")
st.caption("上传叶片照片AI 帮你识别病害")
st.divider()
st.subheader("📷 上传图片")
uploaded_file = st.file_uploader(
"选择香蕉叶片图片",
type=["jpg", "jpeg", "png", "webp"],
help="支持 JPG、PNG、WebP 格式",
)
st.divider()
st.subheader(" 使用说明")
st.markdown("""
1. 上传一张香蕉叶片照片
2. 系统自动识别病害类型
3. 查看各类别置信度
4. 获取防治建议
**Tips:**
- 图片越清晰,识别越准确
- 尽量拍摄完整叶片
- 病变部位清晰可见
""")
st.divider()
st.subheader("📋 可识别病害")
for label_key, info in DISEASE_INFO.items():
st.markdown(
f"<span style='color:{info['color']}'>●</span> **{info['name']}** ({info['name_en']})",
unsafe_allow_html=True,
)
# ─── Main Content ───────────────────────────────────────────────────────────
st.title("🍌 香蕉叶病害智能识别系统")
st.caption("基于 ViT 视觉变换器 · 7 类香蕉叶病害自动分类")
# 加载模型
with st.spinner("正在加载模型(首次运行需下载约 343MB..."):
classifier = load_model()
if uploaded_file is not None:
# 显示上传的图片
image = Image.open(uploaded_file).convert("RGB")
col_img, col_result = st.columns([1, 1])
with col_img:
st.subheader("🖼️ 上传图片")
st.image(image, use_container_width=True)
# 推理
with st.spinner("正在识别病害..."):
results = classifier(image)
# 解析结果
top_result = results[0]
top_label = top_result["label"]
top_score = top_result["score"]
disease = DISEASE_INFO.get(top_label, {
"name": top_label,
"name_en": top_label,
"color": "#95a5a6",
"description": "未知病害类型",
"advice": "建议咨询专业植保人员",
})
with col_result:
st.subheader("🔍 识别结果")
is_healthy = top_label == "LABEL_3"
if is_healthy:
st.success(f"**{disease['name']}** — 置信度 {top_score:.1%}")
else:
st.warning(f"**{disease['name']}** — 置信度 {top_score:.1%}")
st.markdown(f"**英文名**: {disease['name_en']}")
st.markdown(f"**症状描述**: {disease['description']}")
# ─── Confidence Bar Chart ────────────────────────────────────────────────
st.subheader("📊 各类别置信度")
labels = [r["label"] for r in results]
scores = [r["score"] for r in results]
display_names = [
f"{DISEASE_INFO.get(l, {}).get('name', l)}" for l in labels
]
colors = [
DISEASE_INFO.get(l, {}).get("color", "#95a5a6") for l in labels
]
fig = go.Figure()
fig.add_trace(go.Bar(
x=scores,
y=display_names,
orientation="h",
marker=dict(
color=colors,
opacity=0.85,
line=dict(color="rgba(0,0,0,0.08)", width=1),
),
text=[f"{s:.1%}" for s in scores],
textposition="outside",
textfont=dict(color="#5a5a5a", size=11),
))
fig.update_layout(
xaxis=dict(
title="置信度",
range=[0, 1.15],
tickformat=".0%",
color="#5a5a5a",
gridcolor="rgba(0,0,0,0.06)",
),
yaxis=dict(color="#2c2c2c", autorange="reversed"),
paper_bgcolor="rgba(0,0,0,0)",
plot_bgcolor="rgba(0,0,0,0)",
font=dict(color="#2c2c2c", size=11),
margin=dict(t=10, b=30, l=130, r=60),
height=320,
showlegend=False,
)
st.plotly_chart(fig, use_container_width=True)
# ─── Pie Chart ──────────────────────────────────────────────────────────
col_pie, col_advice = st.columns([1, 1])
with col_pie:
st.subheader("📈 置信度分布")
fig_pie = go.Figure()
fig_pie.add_trace(go.Pie(
labels=display_names,
values=scores,
marker=dict(colors=colors),
textinfo="label+percent",
textfont=dict(size=10),
hole=0.45,
))
fig_pie.update_layout(
paper_bgcolor="rgba(0,0,0,0)",
font=dict(color="#2c2c2c", size=10),
margin=dict(t=10, b=10, l=10, r=10),
height=300,
showlegend=False,
)
st.plotly_chart(fig_pie, use_container_width=True)
# ─── Disease Advice ─────────────────────────────────────────────────────
with col_advice:
st.subheader("💡 防治建议")
if is_healthy:
st.success(f"叶片状态健康,继续保持良好的田间管理!")
else:
st.error(f"检测到 **{disease['name']}** 病害")
st.info(disease["advice"])
# 如果置信度不高,给出提示
if top_score < 0.6:
st.warning(
"⚠️ 置信度较低,建议:\n"
"- 确保图片清晰且病变部位可见\n"
"- 尝试拍摄完整叶片\n"
"- 可咨询专业植保人员确认"
)
else:
# 无图片上传时的占位内容
st.info("👈 请在左侧上传一张香蕉叶片图片开始识别")
# 展示示例区域
st.subheader("📖 支持识别的病害类型")
cols = st.columns(4)
for i, (label_key, info) in enumerate(DISEASE_INFO.items()):
with cols[i % 4]:
with st.container(border=True):
st.markdown(
f"<div style='color:{info['color']}; font-size:24px; font-weight:bold;'>●</div>",
unsafe_allow_html=True,
)
st.markdown(f"**{info['name']}**")
st.caption(info["name_en"])
st.divider()
st.caption("香蕉叶病害智能识别系统 · 基于 ViT (Vision Transformer) · 仅供参考,确诊请咨询专业植保人员")