- "Soybean pod and seed counting in both outdoor fields and indoor laboratories using unions of deep neural networks", 2024 ECCV workshop
Outdoor soybean images can be downloaded here P2PNet-Soy.
Some other images used in this study that not taken by us can be downloaded here YOLO-Pod.
The soybean images and detection-annotation files in our paper can be downloaded here Normal dataset, Domain adaptation dataset.
You may need to place the images from other studies into the downloaded dataset's right image folder (corresponding to the annotation files in the training and evaluation sets).
Corresponding annotation files can be find in this repository (yolo_soybean/datasets).
Use YOLOv8_for_soybean.ipynb, YOLOv8_SAM.ipynb, YOLOv8_DA_for_soybean.ipynb at /yolo_soybean/ultralytics/ to train and inference the YOLOv8, YOLOv8-SAM, YOLOv8-DA, respectively.
Pretrained checkpoints can be downloaded here Dropbox.
Specifically, high-quality segment-anything model can be downloaded from Github HQ-SAM.
For indoor segmentation and counting model, we have built a much more lightweight model and it is coming soon.
The real soybean images and instance segmentation annotations can be downloaded here Google Drive.
You can use Simulation/main.py to generate more data for training.
The Mask-RCNN checkpoints can be downloaded here Dropbox.
The Swin-Transformer checkpoints can be downloaded here Dropbox.