xprize-rainforest

This project is a pipeline for processing geospatial data, specifically for the purpose of detecting and segmenting trees in a forest. The pipeline includes several tasks such as tiling, detection, aggregation, and segmentation. Each task is configurable via YAML configuration files.

Installation

Clone the repository to your local machine:

git clone https://github.com/hugobaudchon/xprize-rainforest.git
cd pipelines-rainforest

Install the required Python packages:

pip install -r requirements.txt

Create a dotenv file and the local path to your root directory

touch .env
echo "LOCAL_PATH=/your/local/path" > .env

Configuration

The pipeline is configured using YAML files. Sample configuration files are provided in the config/samples directory. You can copy these files and modify them according to your needs.

Each configuration file has sections for different tasks. For example, the xprize.yaml file includes sections for tilerizer, detector, aggregator, and segmenter tasks. Each section contains parameters specific to that task.

Usage

The main entry point of the pipeline is main.py. This script accepts command-line arguments specifying the task and subtask to perform, and the path to the configuration file.

Here is a basic usage example:

python main.py --task pipelines --config_path ./config/pipelines.yaml

Tasks and Subtasks

The following table lists the available tasks and subtasks:

Task	Subtask	Description	Config Sample
pipeline	xprize	Runs the entire inference pipeline: pipeline-detector, pipeline-segmenter, pipeline-classifier.	pipeline_xprize_sample.yaml
pipeline	detector	Runs the detection inference pipeline: tilerizer, detector, aggregator, coco_to_geopackage.	pipeline_detector_sample.yaml
pipeline	segmenter	Runs the segmentation inference pipeline: tilerizer, segmenter, aggregator, and coco_to_geopackage.	pipeline_segmenter_sample.yaml
pipeline	classifier	Runs the classification inference pipeline: tiling, embedder, classifier.	pipeline_classifier_sample.yaml
tilerizer	N/A	Splits the input raster into tiles.	tilerizer_sample.yaml
detector	train	Trains the detector model.	detector_sample.yaml
detector	score	Scores the detector model.	detector_sample.yaml
detector	infer	Runs inference using the detector model.	detector_sample.yaml
aggregator	N/A	Aggregates the detection results.	aggregator_sample.yaml
segmenter	infer	Runs inference using the segmenter model.	segmenter_sample.yaml

For each task, you need to provide a configuration file with the appropriate sections filled out. For example, to run the detector 'train' subtask, you would use a command like this:

python main.py --task detector --subtask train --config_path ./config/detector.yaml

Please refer to the sample configuration files in the config/samples directory for examples of how to set up the configuration files.