Python FeatureExtractor.transform方法代码示例

# 需要导入模块: from feature_extractor import FeatureExtractor [as 别名]
# 或者: from feature_extractor.FeatureExtractor import transform [as 别名]
def main():
    dataset_path = "/path/to/Caltech-101"
    modelzoo_path = "/path/to/VGG16"
    
    # create an instance
    convnet = FeatureExtractor(
            prototxt_path=os.path.join(modelzoo_path, "vgg16_deploy.prototxt"),
            caffemodel_path=os.path.join(modelzoo_path, "vgg16.caffemodel"),
            target_layer_name="fc7",
            image_size=224,
            mean_values=[103.939, 116.779, 123.68])
    
    # header
    f = open("caltech101_vggnet_fc7_features.csv", "w")
    header = ["filepath"]
    for i in xrange(4096):
        header.append("feat%d" % (i+1))
    header = ",".join(header) + "\n"
    f.write(header)
    
    # extract features
    categories = os.listdir(dataset_path)
    for category in pyprind.prog_bar(categories):
        file_names = os.listdir(os.path.join(dataset_path, category))
        for file_name in file_names:
            img = cv2.imread(os.path.join(dataset_path, category, file_name))
            feat = convnet.transform(img)
            feat_str = [os.path.join(category, file_name)]
            for value in feat:
                feat_str.append(str(value))
            row = ",".join(feat_str)
            f.write("%s\n" % row)
            f.flush()

    f.close()

# 需要导入模块: from feature_extractor import FeatureExtractor [as 别名]
# 或者: from feature_extractor.FeatureExtractor import transform [as 别名]
def main():
    caffe_alexnet_path = "/path/to/caffe-modelzoo/AlexNet"
    caffe_vgg16_path = "/path/to/caffe-modelzoo/VGG16"
    caffe_googlenet_path = "/path/to/caffe-modelzoo/GoogleNet"
    keys_path = "/path/to/dataset/keys.txt"
    data_path = "/path/to/dataset/images"
    dst_path = "/path/to/dataset/features.npy"

    modelname = "VGG16"

    # load pre-trained model
    if modelname == "AlexNet":
        if not os.path.exists(os.path.join(caffe_alexnet_path, "imagenet_mean.npy")):
            convert_mean_file(caffe_alexnet_path)
        convnet = FeatureExtractor(
                prototxt_path=os.path.join(caffe_alexnet_path, "alexnet_deploy.prototxt"),
                caffemodel_path=os.path.join(caffe_alexnet_path, "alexnet.caffemodel"),
                target_layer_name="fc6",
                image_size=227,
                mean_path=os.path.join(caffe_alexnet_path, "imagenet_mean.npy")
                )
    elif modelname == "VGG16":
        convnet = FeatureExtractor(
                prototxt_path=os.path.join(caffe_vgg16_path, "vgg16_deploy.prototxt"),
                caffemodel_path=os.path.join(caffe_vgg16_path, "vgg16.caffemodel"),
                target_layer_name="fc6",
                image_size=224,
                mean_values=[103.939, 116.779, 123.68]
                )
    elif modelname == "GoogleNet":
        googlenet = FeatureExtractor(
                prototxt_path=os.path.join(caffe_googlenet_path, "googlenet_deploy.prototxt"),
                caffemodel_path=os.path.join(caffe_googlenet_path, "googlenet.caffemodel"),
                target_layer_name="pool5/7x7_s1",
                image_size=224,
                mean_values=[104.0, 117.0, 123.0]
                )
    else:
        print "Unknown model name: %s" % modelname
        sys.exit(-1)
    
    # data list
    keys = load_keys(keys_path)
    
    # feature extraction
    feats = []
    for key in keys:
        img = cv2.imread(os.path.join(data_path, key))
        assert img is not None
        feat = convnet.transform(img)
        feats.append(feat)
    feats = np.asarray(feats)
    np.save(dst_path, feats)

    print "Done."

# 需要导入模块: from feature_extractor import FeatureExtractor [as 别名]
# 或者: from feature_extractor.FeatureExtractor import transform [as 别名]
def train_model(X_df, y_array, skf_is):
    fe = FeatureExtractor()
    fe.fit(X_df, y_array)
    X_array = fe.transform(X_df)
    # Regression
    train_is, _ = skf_is
    X_train_array = np.array([X_array[i] for i in train_is])
    y_train_array = np.array([y_array[i] for i in train_is])
    reg = Regressor()
    reg.fit(X_train_array, y_train_array)
    return fe, reg

# 需要导入模块: from feature_extractor import FeatureExtractor [as 别名]
# 或者: from feature_extractor.FeatureExtractor import transform [as 别名]
df = df.drop(df_tmp.index)

from regressor import Regressor
from feature_extractor import FeatureExtractor

df_features = df.drop('target', axis=1)
y = df.target.values

df_train, df_test, y_train, y_test = train_test_split(df_features, y, test_size=0.5, random_state=42)


feature_extractor = FeatureExtractor()
model = Regressor()


X_train = feature_extractor.transform(df_train)
model.fit(X_train, y_train)

X_test = feature_extractor.transform(df_test)
y_pred = model.predict(X_test)
print('RMSE = ', np.sqrt(mean_squared_error(y_test, y_pred)))


imputer = model.clf.named_steps['imputer']

valid_idx = imputer.transform(np.arange(df_train.shape[1])).astype(np.int)
et = model.clf.named_steps['extratreesregressor']

feature_importances = pd.DataFrame(data=et.feature_importances_,
                                   index=df_train.columns[valid_idx][0])
feature_importances['counts'] = df_train.count()[valid_idx][0]