DetectionEvaluationJ is an ImageJ plugin that has been designed to evaluate the performance of object detection algorithms using several metrics. From a set of images, the gold standard associated with such images, and the detected regions obtained by the detection algorithm, DetectionEvaluationJ generates a report that summarises the performance of the detection algorithm based on the available measures.

DetectionEvaluationJ is free to use and licensed under the license GNU GPL 3.0.

Download

You can download DetectionEvaluationJ from here.

Dependencies

ImageJ or Fiji.

Installation

  1. Download and install either ImageJ or Fiji.
  2. Download the DetectionEvaluationJ.
  3. Put the jar file downloaded in the previous step into your ImageJ/Fiji plugins folder.

After restarting ImageJ/Fiji, you will find the option IJ-OpenCV under the plugins menu. That submenu contains the plugins developed with this library.

Usage example

  1. Open an image in ImageJ or Fiji.
  2. From the Plugins menu select the option DetectionEvaluationJ. The following window will be shown.
  3. Select a region (or regions) in the image and press the "Add Gold Standard Roi" button. The selected regions can be saved by pressing the "Save Gold Standard Roi" button, and loaded later using the "Load Gold Standard Roi".
  4. Select a region that will play the role of the region detected by the algorithm. The detected regions can be loaded using the "Load Hypothesised Roi" button.
  5. Press the "Performance Evaluation" button.
  6. Select the matchings between the gold standard and the detected regions in the window that appears.

    If you want to compare several algorithms mark the "Add results to previous data" checkbox. Press the "Evaluate" button.
  7. The measurements and the ROC space are shown.

Source files

The source code of DetectionEvaluationJ is also freely available at the following github page.

Contact

Jónathan Heras joheras@gmail.com.