https://github.com/Ye-Sk/TasselLFANet

TasselLFANet：Lightweight Feature Aggregation Neural Network
https://github.com/Ye-Sk/TasselLFANet

Last synced: 7 days ago
JSON representation

TasselLFANet：Lightweight Feature Aggregation Neural Network

• Host: GitHub
• URL: https://github.com/Ye-Sk/TasselLFANet
• Owner: Ye-Sk
• Created: 2023-01-26T13:20:03.000Z (almost 2 years ago)
• Default Branch: master
• Last Pushed: 2024-02-10T02:05:06.000Z (9 months ago)
• Last Synced: 2024-08-02T15:29:53.907Z (3 months ago)
• Language: Python
• Homepage:
• Size: 3.55 MB
• Stars: 3
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# TasselLFANet



  

  


TasselLFANet based on detection method for plant counting, implementation of paper :   

[___TasselLFANet：A Novel Lightweight Multi-Branch Feature Aggregation Neural Network for High-throughput Image-based Maize Tassels Detection and Counting___](https://www.frontiersin.org/articles/10.3389/fpls.2023.1158940/abstract)

# Main results

### Object Detection on MrMT dataset

|Model|FPS|P|R|F1|[email protected]|[email protected]:.95|

| :----: | :----: | :----: | :----: | :----: | :----: | :----: |

|LFANet-HE|125|0.947|0.926|0.936|0.962|0.518|

|LFANet|77|0.946|0.942|0.944|0.968|0.546|

* Speed is tested on Nvidia Quadro P5000 GPU（16G）

### Object Counting on MrMT dataset

|Model|MAE|RMSE|MAPE|R²|

| :----: | :----: | :----: | :----: | :----: |

|LFANet-HE|2.70|3.76|14.3%|0.9751|0.9751|

|LFANet|1.80|2.68|9.2%|0.9903|0.9903|

# Installation

1. The code we implement is based on PyTorch 1.8 and Python 3.6, please refer to the file `requirements.txt` to configure the required environment.      

2. To convenient install the required environment dependencies, you can also use the following command look like this :     

~~~

$ pip install -r requirements.txt 

~~~

# Build your own dataset

**To train your own datasets on this framework, we recommend that :**  

* Annotate your data with the image annotation tool [LabelIMG](https://github.com/heartexlabs/labelImg) to generate `.txt` labels.   

* Refer to the `config/data.yaml` example to configure your own hyperparameters file. 

* Based on the `train.py` code example configure your own training parameters.

# Training

### Prepare Your Data

1. You can download the `MrMT` dataset from [___Baidu Drive (9.2GB)___](https://github.com/Ye-Sk/MrMT)

2. Move your dataset into the `data` folder, please follow the format look like this :

~~~

├── data

│   ├── images

│   │   ├── train

│   │   ├── valid

│   │   └── test

│   ├── labels

│   │   ├── train

│   │   ├── valid

│   │   └── test

~~~

* Run the following command to start training :  

~~~

$ python train.py --dataset config/data.yaml --batch-size 16 --workers 8

~~~

___For some reasons, our experiment haven't use a pretrained model, and we recommend 

that you pretrain if resources are adequate, the gains from this are considerable.___

# Evaluation

* Run the following command to evaluate the results :  

~~~

$ python eval.py --model LFANet.pt --dataset config/data.yaml --imgsz 640

~~~

# Inference

* Run the following command on a variety of sources :   

~~~

$ python infer.py --imgsz 640 --source config/images  # on image

~~~

~~~

$ python infer.py --imgsz 640 --source 0  # on webcam

~~~

# Citation

~~~

@article{ye2023TasselLFANet,  

  title={TasselLFANet: A Novel Lightweight Multi-Branch Feature Aggregation Neural Network for High-throughput Image-based Maize Tassels Detection and Counting},  

  author={Yu, Zhenghong and Ye, Jianxiong and Li, Cuina and Zhou, Huabing and Li, Xun}, 

  journal={Frontiers in Plant Science}, 

  volume={14},

  pages={1291-1307},

  year={2023},

  doi={10.3389/fpls.2023.1158940}

}

~~~