diff --git a/README.md b/README.md index 2930d9d..22043e9 100644 --- a/README.md +++ b/README.md @@ -43,7 +43,9 @@ Then, follow the pipeline! ```bash python3 dataset.py ``` -This will create a `tiles` folder containing the tiles of the processed satellite images. The path to the processed satellite images that are to be processed should be updated in the `config.yaml` file. +Because the satellite images are too big to be processed efficiently by `detectron2` we split the images into `tiles`. **Tile size** can be changed in the `config.yaml` file and can slighlty change the results (we recommand to do a `hyperparameter search` to search for the optimal tile size). + +The script will create a `tiles` folder containing the tiles of the processed satellite images. The path to the processed satellite images that are to be processed should be updated in the `config.yaml` file. ## Train detectron2 @@ -51,6 +53,8 @@ This will create a `tiles` folder containing the tiles of the processed satellit python3 train.py ``` +With this script we train a `Detectron2` model with a [faster-rcnn architecture](https://github.com/facebookresearch/detectron2/blob/main/configs/COCO-Detection/fast_rcnn_R_50_FPN_1x.yaml) as backbone. Note that it is possible to change the backbone by choosing another [Detectron compliant](https://github.com/facebookresearch/detectron2/tree/main/configs/COCO-Detection) model. + The script should create a folder `output` that contains `model_final.pth`, the logs and other files that `Detectron2` creates. ## Predict