SHAP-Based Interpretable Object Detection Method for Satellite Imagery
July 29, 2022 ยท View on GitHub
This is the author implementation of SHAP-Based Interpretable Object Detection Method for Satellite Imagery. The implementation of the object detection model (YOLOv3) is based on Pytorch_YOLOv3. The framework of the proposed method can be applied to any differentiable object detection model.
Performance
Visualization
Please see the paper for details on the results of the evaluation, regularization, and data selection methods.
Installation
Requirements
- Python 3.6.3+
- Numpy
- OpenCV
- Matplotlib
- Pytorch 1.2+
- Cython
- Cuda (verified as operable: v10.2)
- Captum (verified as operable: v0.4.1)
optional:
- tensorboard
- tensorboardX
- CuDNN
Download the original YOLOv3 weights
download the pretrained file from the author's project page:
$ mkdir weights
$ cd weights/
$ bash ../requirements/download_weights.sh
Usage
Please see the test.ipynb
Paper
SHAP-based Methods for Interpretable Object Detection in Satellite Imagery
Hiroki Kawauchi, Takashi Fuse