Confusion Matrix for Object Detection and Instance Segmentation

metya 1dbb315ae8 Refactor a little bit vor 3 Jahren
.gitignore 25f6cf35fc First Commit with ConfusionMatrix class vor 3 Jahren
README.md 25f6cf35fc First Commit with ConfusionMatrix class vor 3 Jahren
confusion_matrix.py 1dbb315ae8 Refactor a little bit vor 3 Jahren
setup.py 25f6cf35fc First Commit with ConfusionMatrix class vor 3 Jahren

README.md

Confusion Matrix for Object Detection and Instance Segmentation

That's my implementation of class aware confusion matrix for object detection and instance segmentations. Particulary it uses COCO format of datasets for targets and predictions. But it easily can be rewrited to another format type. Also it uses pytorch for typings, but again easily can be replaces with tensorflow for example.

Using

>>> from confusion_matrix import ConfusionMatrix

>>> confusion_matrix = ConfusionMatrix(class_names={0: 'class1', 1: 'class2'},
...                                            thrs_config={0: 0.5, 1: 0.5})
>>> for images, targets in test_dataloader:
>>>     outputs = model(images)
>>>     confusion_matrix.update(outputs, targets)

>>> confusion_matrix.plot(show=True)

or 

>>> confusion_matrix.pretty_plot()

Instalation

pip install git+https://github.com/metya/confusion_matrix