- In
tools/, we provide a series of handy scripts for converting data formats and training the models. - In
scripts/, it lists specific command for running the code for processing the given dataset. - The
configs/contains the configuration for different deep learning models, and is organized by datasets.
- Get the dataset and annotations -- if you are not sure, feel free to check this tutorial.
- Duplicate and modify the config files and training scripts
- For example, you might want to copy
configs/prima/fast_rcnn_R_50_FPN_3xtoconfigs/your-dataset-name/fast_rcnn_R_50_FPN_3x, and you can create your ownscripts/train_<your-dataset-name>.shbased onscripts/train_prima.sh. - You'll modify the
--dataset_name,--json_annotation_train,--image_path_train,--json_annotation_val,--image_path_val, and--config-fileargs appropriately.
- For example, you might want to copy
- If you have a dataset with segmentation masks, you can try to train with the
mask_rcnn model; otherwise you might want to start with thefast_rcnn model- If you see error
AttributeError: Cannot find field 'gt_masks' in the given Instances!during training, this means you should not use
- If you see error
- Prima Layout Analysis Dataset
scripts/train_prima.sh- You will need to download the dataset from the official website and put it in the
data/primafolder. - As the original dataset is stored in the PAGE format, the script will use
tools/convert_prima_to_coco.pyto convert it to COCO format. - The final dataset folder structure should look like:
data/ └── prima/ ├── Images/ ├── XML/ ├── License.txt └── annotations*.json
- You will need to download the dataset from the official website and put it in the
- cocosplit A script that splits the coco annotations into train and test sets.
- Detectron2 Detectron2 is Facebook AI Research's next generation software system that implements state-of-the-art object detection algorithms.