-
This code is an implementation of our manuscript: A New Deep Learning-based Dynamic Paradigm Towards Open-World Plant Disease Detection
-
The manuscript is currently under peer review. We submitted it to Frontiers in Plant Science.
-
Authors: Jiuqing Dong, Alvaro Fuentes, Sook Yoon*, Hyongsuk Kim*, Yongchae Jeong, Dong Sun Park
-
You can install the environment following this https://github.com/JosephKJ/OWOD
-
Set up environment
conda create -n OWPDD python=3.7
conda activate OWPDD
conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3 -c pytorch -c conda-forge
-
install dependecies
pip install -r requirement.txt
-
install mmdet (will take a while to process)
unzip detectron2.zip
python -m pip install -e ./
- You can use our pre-trained model on google drive to implement Well-trained UA-RPN in our work.
- If you use the Object365 model, please cite our paper. Other pre-trained models are from Detectron2.
- We use Object365 pre-trained model to evaluate the OWOD task on the COCO dataset. We provide a part of the result at COCO.xlsx
- Kindly check run.sh file for a task workflow.
Operational steps for the teacher-student learning paradigm: step 0: Create your dataset, suffix of the annotation file should be '.xml'. If you don/t know how to transfer the Json to XML, please refer this: https://github.com/bot66/coco2voc
step 1: Put the customer dataset into ./datasets, the structure should be like this:
-..
-datasets
-VOC2007
-JPEGImages
-000001.jpg *
-Annotations
-000001.xml *
-ImageSets
-Main
-t1_train.txt ('.txt' file save the filename without suffix， eg. '000001')
-t2_train.txt
-t2_ft.txt
...
-all_task_val.txt
-all_task_test.txt
step 2: Change the config file, such as NUM_CLASSES, OWOD.CUR_INTRODUCED_CLS, PREV_INTRODUCED_CLS, and so on. You also can define your own dataset name. Note that the name matching at './detectron2/data/datasets/builtin.py'
step 3: Change the category names at './detectron2/data/datasets/pascal_voc.py'
step 4: Train the model. Kindly check run.sh file for a task workflow, and then run it. For example:
- sh run.sh