---
title: cv_alex_video-single-object-tracking_siamfc
canonical_url: "https://www.modelscope.cn/models/damo/cv_alex_video-single-object-tracking_siamfc"
md_url: "https://www.modelscope.cn/models/damo/cv_alex_video-single-object-tracking_siamfc.md"
repository: damo/cv_alex_video-single-object-tracking_siamfc
chinese_name: "Siamfc视频单目标跟踪-通用领域-S"
last_updated: 2023-02-09
license: "MIT License"
pipeline_tag: video-single-object-tracking
tasks:
  - video-single-object-tracking
model_type:
  - SiamfcTrack
library_name:
  - pytorch
frameworks:
  - pytorch
domain:
  - cv
downloads: 3782
stars: 2
tags:
  - transformer
  - single_object_tracking
  - SiamfcTrack
  - uav
---

# cv_alex_video-single-object-tracking_siamfc

> cv_alex_video-single-object-tracking_siamfc - damo 在 ModelScope 开源的模型。单目标跟踪算法模型介绍 对于一个输入视频，只需在第一帧图像中用矩形框指定待跟踪目标，单目跟踪算法将在整个视频帧中持续跟踪该目标，输出跟踪目标在所有图像帧中的矩形框信息。

damo/cv_alex_video-single-object-tracking_siamfc 是 ModelScope 魔搭社区上的video-single-object-tracking模型，采用 MIT License 许可。

- **Repository**: damo/cv_alex_video-single-object-tracking_siamfc
- **License**: MIT License
- **Tasks**: video-single-object-tracking
- **Tags**: transformer, single_object_tracking, SiamfcTrack, uav
- **Downloads**: 3782
- **Stars**: 2
- **Last updated**: 2023-02-09

Source: https://www.modelscope.cn/models/damo/cv_alex_video-single-object-tracking_siamfc

---

<!--- 以下model card模型说明部分，请使用中文提供（除了代码，bibtex等部分） --->

# <OSTrack>单目标跟踪算法模型介绍
对于一个输入视频，只需在第一帧图像中用矩形框指定待跟踪目标，单目跟踪算法将在整个视频帧中持续跟踪该目标，输出跟踪目标在所有图像帧中的矩形框信息。




## 模型描述
<img src="resources/intro.jpg" width="800" >

本模型是基于Siamfc方案的单目标跟踪框架，使用AlexNet作为主干网络进行训练，是One-Stream单目标跟踪算法。

## 期望模型使用方式以及适用范围

该模型适用于视频单目标跟踪场景。

### 如何使用模型

- 根据输入待跟踪视频和第一帧图像对应的待跟踪矩形框（x1, y1, x2, y2），可按照代码范例进行模型推理和可视化。

#### Installation
```
conda create -n anti_uav python=3.7
conda activate anti_uav
# pytorch >= 1.3.0
pip install torch==1.8.1+cu102  torchvision==0.9.1+cu102 torchaudio==0.8.1  --extra-index-url https://download.pytorch.org/whl/cu102
git clone https://github.com/ly19965/CVPR_Anti_UAV
cd CVPR_Anti_UAV
pip install -r requirements/tests.txt 
pip install -r requirements/framework.txt
pip install -r requirements/cv.txt 
```

#### 推理代码范例
```python
from modelscope.utils.cv.image_utils import show_video_tracking_result
from modelscope.outputs import OutputKeys
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks

video_single_object_tracking = pipeline(Tasks.video_single_object_tracking, model='damo/cv_alex_video-single-object-tracking_siamfc')
video_path = "https://modelscope.oss-cn-beijing.aliyuncs.com/test/videos/dog.avi"
init_bbox = [414, 343, 514, 449] # the initial object bounding box in the first frame [x1, y1, x2, y2]
result = video_single_object_tracking((video_path, init_bbox))
show_video_tracking_result(video_path, result[OutputKeys.BOXES], "./tracking_result.avi")
print("result is : ", result[OutputKeys.BOXES])
```

#### Multi-GPU模型训练代码范例

```python
import os.path as osp
from modelscope.metainfo import Trainers
from modelscope.trainers import build_trainer
from modelscope.msdatasets import MsDataset
from modelscope.utils.constant import DownloadMode
import os
import json

# Step 1: 数据集准备
train_dataset = MsDataset.load('Got-10k', namespace='ly261666', split='train')

# Step 2: 相关参数设置
data_root_dir = '/home/ly261666/workspace/maas/modelscope_project/Mass_env/or_data/train_data/got10k_v1' # 下载的数据集路径
model_id = 'damo/cv_alex_video-single-object-tracking_siamfc'
cache_path = '/home/ly261666/.cache/modelscope/hub/damo/cv_alex_video-single-object-tracking_siamfc'# 下载的modelscope模型路径
cfg_file = os.path.join(cache_path, 'configuration.json')

kwargs = dict(
    cfg_file=cfg_file,
    model=model_id, # 使用DAMO-YOLO-S模型 
    gpu_ids=[  # 指定训练使用的gpu
    0,1,2,3,4,5,6,7
        ],
    batch_size=64, # batch_size, 每个gpu上的图片数等于batch_size // len(gpu_ids)
    max_epochs=300, # 总的训练epochs
    load_pretrain=False, # 是否载入预训练模型，若为False，则为从头重新训练, 若为True，则加载modelscope上的模型finetune。
    base_lr_per_img=0.001, # 每张图片的学习率，lr=base_lr_per_img*batch_size
    train_image_dir=data_root_dir, # 训练图片路径
    val_image_dir=data_root_dir, # 测试图片路径
    )


# Step 3: 开启训练任务
if __name__ == '__main__':
    trainer = build_trainer(
                        name=Trainers.video_single_object_tracking, default_args=kwargs)
    trainer.train()

```

### 模型局限性以及可能的偏差
- 在遮挡严重场景和背景中存在与目标高度相似的物体场景下，目标跟踪精度可能欠佳。
- 建议在有GPU的机器上进行测试，由于硬件精度影响，CPU上的结果会和GPU上的结果略有差异。


### 相关论文以及引用信息
本模型主要参考论文如下：

```BibTeX
@inproceedings{bertinetto2016fully,
      title={Fully-convolutional siamese networks for object tracking},
        author={Bertinetto, Luca and Valmadre, Jack and Henriques, Joao F and Vedaldi, Andrea and Torr, Philip HS},
          booktitle={Computer Vision--ECCV 2016 Workshops: Amsterdam, The Netherlands, October 8-10 and 15-16, 2016, Proceedings, Part II 14},
            pages={850--865},
              year={2016},
                organization={Springer}
}
```