pcr-framework 2026.10.22

Description

Modular computer vision library developed by the PCR-UPM group for high-performance image processing tasks. Designed with research and engineering in mind, the framework provides reusable components for image preprocessing, and model evaluation. It is built on top of widely adopted libraries such as OpenCV, NumPy, and Rasterio.

System Requirements

This package requires GDAL. It is recommended to install it using Conda:

conda install -c conda-forge gdal

Requisites

• pcr_framework https://github.com/pcr-upm/pcr_framework
• tqdm
• scikit-learn (training)
• natsort (evaluation)

Installation

This repository must be located before the following directory:

images_experiments
    └── pcr_framework

Usage

usage: images_experiments.py [-h] --mode {train,test} --part
                             {detection,segmentation,recognition,all}
                             --anns-file ANNS_FILE [--show-viewer]
                             [--save-file] [--save-image] --detection
                             {retinanet17_detection,scrdet19_detection} 
                             --segmentation
                             {unet15_segmentation,ocr20_segmentation}
                             --recognition
                             {resnet15_recognition}

• Use the -h option to show the help message and exit.

• Use the --mode option to choose the mode (train, test).

• Use the --part option to choose the tasks (detection, segmentation, recognition, all).

• Use the --anns-file option to define the annotations input file.

• Use the --show-viewer option to show results visually.

• Use the --save-file option to save experiments in a text file.

• Use the --save-image option to save the processed images.

• Use the --detection option to select the detection module.

• Use the --segmentation option to select the segmentation module.

• Use the --recognition option to select the recognition module.

> python test/images_experiments.py -m train -p detection -a database/detection/15cm/cowc/cowc_ann_train_train.txt -d scrdet19_detection --database COWC --gpu 0 --gpu 1

> python test/images_experiments.py -m test -p detection -a database/detection/15cm/cowc/cowc_ann_train_test.txt -d scrdet19_detection --database COWC --gpu 0

> python test/images_experiments.py -m test -p all -a database/detection/15cm/cowc/cowc_ann_train_test.txt -d scrdet19_detection -s unet15_segmentation -r resnet15_recognition --database COWC --gpu 0