PDT Dataset
收藏资源简介:
PDT数据集是由山东计算机科学中心(国家超级计算济南中心)和齐鲁工业大学(山东省科学院)联合开发的无人机目标检测数据集,专门用于检测树木病虫害。该数据集包含高分辨率和低分辨率两种版本,共计5775张图像,涵盖了健康和受病虫害影响的松树图像。数据集的创建过程包括实地采集、数据预处理和人工标注,旨在为无人机在农业中的精准喷洒提供高精度的目标检测支持。PDT数据集的应用领域主要集中在农业无人机技术,旨在提高无人机在植物保护中的目标识别精度,解决传统检测模型在实际应用中的不足。
The PDT dataset is a UAV target detection dataset jointly developed by Shandong Computer Science Center (National Supercomputing Center in Jinan) and Qilu University of Technology (Shandong Academy of Sciences), specifically designed for detecting tree pests and diseases. It includes two versions: high-resolution and low-resolution, with a total of 5775 images covering both healthy pine trees and those affected by pests and diseases. The dataset was created through three stages: field data collection, data preprocessing and manual annotation, aiming to provide high-precision target detection support for UAV-based precision spraying in agricultural applications. The main application scope of the PDT dataset is agricultural UAV technology, which aims to improve the target recognition accuracy of UAVs in plant protection and address the limitations of traditional detection models in practical applications.
PDT: Uav Target Detection Dataset for Pests and Diseases Tree
数据集概述
该数据集旨在通过无人机视觉识别技术,检测作物中的害虫和疾病。数据集包括两个主要部分:PDT数据集和CWC数据集。
PDT数据集
- 类别: 不健康(unhealthy)
- 图像示例:
- (a) 健康目标
- (b) 不健康目标
- 双分辨率:
- LL: 640×640
- LH: 5472×3648
- 数据集结构:
| 版本 | 类别 | 结构 | 目标图像 | 非目标图像 | 图像尺寸 | 实例数 | 目标数量 (S, M, L) |
|---|---|---|---|---|---|---|---|
| 样本 | 不健康 | 训练 | 81 | 1 | 640×640 | 2569 | 1896, 548, 179 |
| 验证 | 19 | 1 | 640×640 | 691 | 528, 138, 25 | ||
| LL | 不健康 | 训练 | 3166 | 1370 | 640×640 | 90290 | 70418, 16342, 3530 |
| 验证 | 395 | 172 | 640×640 | 12523 | 9926, 2165, 432 | ||
| 测试 | 390 | 177 | 640×640 | 11494 | 8949, 2095, 450 | ||
| LH | 不健康 | - | 105 | 0 | 5472×3648 | 93474 | 93474, 0, 0 |
CWC数据集
- 类别:
- bluegrass, chenopodium_album, cirsium_setosum, corn, sedge, cotton, nightshade, tomato, velvet, lettuce, radish
- 数据集来源:
| 数据集 | 来源 | 类别 | 数量 | 图像尺寸 |
|---|---|---|---|---|
| Corn weed datasets | Corn weed datasets | bluegrass, corn, sedge, chenopodium_album, cirsium_setosum | 250 | 800×600 |
| lettuce weed datasets | lettuce weed datasets | lettuce | 200 | 800×600 |
| radish weed datasets | radish weed datasets | radish | 201 | 800×600 |
| Fresh-weed-data | Fresh-weed-data | nightshade, tomato, cotton, velvet | 115, 116, 24, 38 | 800×600, 586×444, 643×500 |
- 数据集结构:
| 类别 | 结构 | bluegrass | chenopodium_album | cirsium_setosum | corn | sedge | lettuce | radish | nightshade | tomato | cotton | velvet |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 目标图像 | 训练 | 200 | 200 | 200 | 200 | 200 | 200 | 200 | 200 | 200 | 200 | 200 |
| 验证 | 40 | 40 | 40 | 40 | 40 | 40 | 40 | 40 | 40 | 40 | 40 | |
| 测试 | 10 | 10 | 10 | 10 | 10 | 10 | 10 | 10 | 10 | 10 | 10 | |
| 目标数量 | S | 1 | 0 | 0 | 5 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| M | 0 | 0 | 0 | 9 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
| L | 249 | 250 | 250 | 236 | 250 | 444 | 326 | 250 | 210 | 268 | 248 | |
| 图像尺寸 | 800×600 | 800×600 | 800×600 | 800×600 | 800×600 | 800×600 | 800×600 | 800×600 | 800×600 | 800×600 | 586×444 | 643×500 |
模型
- YOLO-DP模型: 用于高精度目标检测的模型,结合PDT和CWC数据集进行评估。
实验结果
- 数据集验证: 使用不同模型在PDT和CWC数据集上的性能评估。
- 消融实验: 对YOLOv5s模型的不同变体进行性能比较。
可视化研究
- PDT数据集检测结果: 展示了PDT数据集的检测结果。
- CWC数据集训练过程: 展示了CWC数据集的训练损失曲线。
论文
- PDT: Uav Target Detection Dataset for Pests and Diseases Tree. Mingle Zhou, Rui Xing, Delong Han, Zhiyong Qi, Gang Li*. ECCV 2024.

- 1PDT: Uav Target Detection Dataset for Pests and Diseases Tree山东计算机科学中心(国家超级计算济南中心),齐鲁工业大学(山东省科学院) · 2024年



