def test_predict_raster(test_config, predict_tfrecords):
#Create class
mod = Houston2018.AttentionModel(config="conf/houston_config.yml")
#Replace config for testing env
for key, value in test_config.items():
for nested_key, nested_value in value.items():
mod.config[key][nested_key] = nested_value
#Create
mod.create()
mod.read_data()
results = mod.predict_raster(predict_tfrecords, batch_size=2)
predicted_raster = visualize.create_raster(results)
#Equals size of the input raster
assert predicted_raster.shape == (12, 15)
print("Unique labels in ytrue {}, unique labels in y_pred {}".format(
np.unique(np.argmax(y_true, 1)), np.unique(np.argmax(y_pred, 1))))
experiment.log_confusion_matrix(y_true=y_true,
y_predicted=y_pred,
labels=list(class_labels.values()),
title="Confusion Matrix")
#Predict
predict_tfrecords = glob.glob(
"/orange/ewhite/b.weinstein/Houston2018/tfrecords/predict/*.tfrecord")
results = model.predict_raster(predict_tfrecords, batch_size=512)
#predicted classes
print(results.label.unique())
predicted_raster = visualize.create_raster(results)
print(np.unique(predicted_raster))
experiment.log_image(name="Prediction",
image_data=predicted_raster,
image_colormap=visualize.discrete_cmap(20,
base_cmap="jet"))
#Save as tif for resampling
prediction_path = os.path.join(save_dir, "prediction.tif")
predicted_raster = np.expand_dims(predicted_raster, 0)
resample.create_tif(
"/home/b.weinstein/DeepTreeAttention/data/processed/20170218_UH_CASI_S4_NAD83.tif",
filename=prediction_path,
numpy_array=predicted_raster)
filename = resample.resample(prediction_path)
experiment.log_image(name="Resampled Prediction",