This project introduces a method based on Faster RCNN to detect the number of soybean flowers to solve the problem of manual counting of soybean flowers. Help breeders understand the genetic mechanism of flower drop and increase soybean yield. The project code comes from the official website of MxNet, modified to be used for soybean flower count problem. The picture shows the overall flow chart of the experiment.
Please refer to the official installation tutorial MxNet Installation
- python>=3.6.5
- MxNet>=1.6
- Anaconda3 (recommended)
Train a default resnet50_v1b model with Pascal VOC on GPU 0:
python train_faster_rcnn.py --gpus 0
Train a resnet50_v1b model on GPU 0,1,2,3:
python train_faster_rcnn.py --gpus 0,1,2,3 --network resnet50_v1b
Check the supported arguments:
python train_faster_rcnn.py --help
Model evaluation
python eval_fasterrcnn.py
python Predict.py
Download link of soybean flower dataset: https://pan.baidu.com/s/1pwYNLUUkG9ueayUrm6sZBw. Password:le80 . If you want to use this dataset, please add the corresponding reference or contact us.