Written by Boris Ginsburg
Loss layer, which produces Confusion Matrix.
- input : the same as all loss layers
- output: 2D Matrix (Blob)
Sources:
- include/caffe/loss_layers.hpp: added ConfusionMatrix Layer
- src/caffe/layers/confusion_matrix_layer.cpp
- src/caffe/test/test_confusion_matrix_layer.cpp: tests
- src/caffe/proto/caffe.proto: added proto parameters
- src/caffe/solver.cpp: modifed Solver::Test(.) to support layers with 2D output
- plot_conf_matrix.py: visualization
Example:
- examples/cifar10/cifar10_cm_solver.prototxt
- examples/cifar10/cifar10_cm_train_test.prototxt
- examples/cifar10/train_cm.sh