https://github.com/buxihuo/OW-YOLO

Detect known and unknown objects in the open world（具有区分已知与未知能力的全新检测器））
https://github.com/buxihuo/OW-YOLO

autolabel deep-learning object-detection open-world open-world-detection pytorch unknown yolo yolov5 yolov8

Last synced: about 1 month ago
JSON representation

Detect known and unknown objects in the open world（具有区分已知与未知能力的全新检测器））

• Host: GitHub
• URL: https://github.com/buxihuo/OW-YOLO
• Owner: buxihuo
• Created: 2022-03-28T07:53:25.000Z (over 2 years ago)
• Default Branch: ow-yolo-det
• Last Pushed: 2023-02-26T04:13:53.000Z (almost 2 years ago)
• Last Synced: 2024-08-02T01:22:37.349Z (5 months ago)
• Topics: autolabel, deep-learning, object-detection, open-world, open-world-detection, pytorch, unknown, yolo, yolov5, yolov8
• Language: Python
• Homepage:
• Size: 6.12 MB
• Stars: 88
• Watchers: 5
• Forks: 7
• Open Issues: 14
• Metadata Files:
  • Readme: README.md
  • Funding: .github/FUNDING.yml

Awesome Lists containing this project

• awesome-yolo-object-detection - buxihuo/OW-YOLO - YOLO?style=social"/> : Detect known and unknown objects in the open world（具有区分已知与未知能力的全新检测器））. (Applications)
• awesome-yolo-object-detection - buxihuo/OW-YOLO - YOLO?style=social"/> : Detect known and unknown objects in the open world（具有区分已知与未知能力的全新检测器））. (Applications)

README

        
![015](https://user-images.githubusercontent.com/84908793/221391769-2503d415-60c3-4c4b-a432-a4f0fdfa186c.png)

# 快速开始

## 1. 安装

  同yolov5

## 2. 推理

 

 推理示例

```bash

$ python detect.py --source data/images --weights m-obj365.pt --unknownconf 0.45 --conf 0.25 

$ python detect.py --source data/images --weights s-coco.pt --unknownconf 0.25 --conf 0.25

'''

参数解读

unknownconf: 当网络预测的“不知道”分数大于此阈值时预测为不知道，否则输出已知分类。

1）与已知类精度关系：当已知类精度越高时，“不知道”在已知类上发生的情况将越少，预测未知类时可以设定更低的unknownconf而不影响已知类性能。

2）与训练集大小关系：训练集越丰富，预测未知类的能力越强

注：可根据需求调节此参数，需要注意的是由于资源问题，在objects365数据集下训练的模型m-obj365.pt仅在小模型下训练了30轮，精度较低，建议采用较高阈值。

其他参数：

1）非极大值抑制：默认类内NMS（非极大值抑制）iou阈值为0.45，参数为iou；同时进行类间NMS，iou阈值为0.75，后续将提供参数接口。

2）不知道的物体类名：可在detect文件中修改，后续将提供接口。             

'''

```

 

 视频展示

1. [demo1](https://b23.tv/MfpEmAm)

2. [demo2](https://www.bilibili.com/video/BV1Nm4y1P7UW/?share_source=copy_web&vd_source=4f63c00122ad06d30c832c5c6f903637)

## 3. 预训练模型

[s-coco.pt](https://github.com/buxihuo/OW-YOLO/releases/download/0.1/s-coco.pt)


[m-obj365.pt](https://github.com/buxihuo/OW-YOLO/releases/download/0.1/m-obj365.pt)


### coco数据集性能

|Model                       |size
(pixels)|mAPval
0.5:0.95 |mAPval
0.5 |

|---                         |---                  |---                     |--- 

|yolov8-n             |640                  |37.3                        |52.6

|OW-yolov8-n          |640                  |37.9 (only known 38)                        |53.7 (only known 53.9)

|OW-yolov8-n-la       |640                  |                        |

```bash

la : label attention

```

## 4. 后续功能

图像分类、实例分割