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

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

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 

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

• Annotate your data with the image annotation tool 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.

1. You can download the MrMT dataset from Baidu Drive (9.2GB)
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.

• Run the following command to evaluate the results :

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

• 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

@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}
}