#!/usr/bin/python
# EE4 Selected Topics From Computer Vision Coursework
# Vasil Zlatanov, Nunzio Pucci

import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import scikitplot as skplt
import argparse
import logging
from logging import debug
from sklearn.cluster import KMeans
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import ExtraTreesClassifier 
from sklearn.ensemble import RandomTreesEmbedding
from sklearn.metrics import accuracy_score
import time 

parser = argparse.ArgumentParser()
parser.add_argument("-d", "--data", help="Data path", action='store_true', default='data.npz')
parser.add_argument("-c", "--conf_mat", help="Show visual confusion matrix", action='store_true')
parser.add_argument("-k", "--kmean", help="Perform kmean clustering with --kmean cluster centers", type=int, default=0)
parser.add_argument("-l", "--leaves", help="Maximum leaf nodes for RF classifier", type=int, default=256)
parser.add_argument("-e", "--estimators", help="number of estimators to be used", type=int, default=100)
parser.add_argument("-D", "--treedepth", help="depth of trees", type=int, default=5)
parser.add_argument("-v", "--verbose", help="Use verbose output", action='store_true')
parser.add_argument("-t", "--timer", help="Display execution time", action='store_true')
parser.add_argument("-T", "--testmode", help="Testmode", action='store_true')
parser.add_argument("-E", "--embest", help="RandomTreesEmbedding estimators", type=int, default=256)
parser.add_argument("-r", "--randomness", help="Randomness parameter", type=int, default=0)

args = parser.parse_args()
if args.verbose:
    logging.basicConfig(level=logging.DEBUG)

def make_histogram(data, model, args):
    if args.kmean:
        hist_size = args.kmean
    else:
        hist_size = args.embest*args.leaves

    histogram = np.zeros((data.shape[0], data.shape[1],hist_size)) 
    for i in range(data.shape[0]):
        for j in range(data.shape[1]):
            if (args.kmean):
                histogram[i][j] = np.bincount(model.predict(data[i][j].T), minlength=args.kmean)
            else:
                leaves = model.apply(data[i][j].T)
                leaves = np.apply_along_axis(np.bincount, axis=0, arr=leaves, minlength=args.leaves)
                histogram[i][j] = leaves.reshape(hist_size)
    return histogram

def run_model (data, train, test, train_part, args):
    if args.timer:
        start = time.time()
    
    if (args.kmean):
        logging.debug("Computing KMeans with", train_part.shape[0], "keywords")
        kmeans = KMeans(n_clusters=args.kmean, n_init=1, random_state=0).fit(train_part)
        hist_train = make_histogram(train, kmeans, args)
        hist_test = make_histogram(test, kmeans, args)
    else:
        trees = RandomTreesEmbedding(max_leaf_nodes=args.leaves, n_estimators=args.embest, random_state=0).fit(train_part)
        hist_train = make_histogram(train, trees, args)
        hist_test = make_histogram(test, trees, args)
 
    logging.debug("Generating histograms")
    
    logging.debug("Keywords shape", hist_train.shape, "\n")
    logging.debug("Planting trees...")
    if args.randomness:
        clf = RandomForestClassifier(max_features=args.randomness, n_estimators=args.estimators, max_depth=args.treedepth, random_state=0)
    else:
        clf = RandomForestClassifier(n_estimators=args.estimators, max_depth=args.treedepth, random_state=0)
    clf.fit(
            hist_train.reshape((hist_train.shape[0]*hist_train.shape[1], hist_train.shape[2])),
            np.repeat(np.arange(hist_train.shape[0]), hist_train.shape[1]))

    logging.debug("Random forests created")

    test_pred = clf.predict(hist_test.reshape((hist_test.shape[0]*hist_test.shape[1], hist_test.shape[2])))
    test_label = np.repeat(np.arange(hist_test.shape[0]), hist_test.shape[1])
    train_pred = clf.predict(hist_train.reshape((hist_train.shape[0]*hist_train.shape[1], hist_train.shape[2])))
    train_label = np.repeat(np.arange(hist_train.shape[0]), hist_train.shape[1])

    if args.timer:
        end = time.time()
        print("Execution time: ",end - start)
    if args.conf_mat:
        skplt.metrics.plot_confusion_matrix(test_pred, test_label, normalize=True)
        plt.show()

    return accuracy_score(test_pred, test_label)#, accuracy_score(train_pred, train_label), end-start
 


def main():
    data = np.load(args.data)
    train = data['desc_tr']
    test = data['desc_te']
    train_part = data['desc_sel'].T
    logging.debug("Verbose is on")

    if args.testmode:
        acc = np.zeros((3,100))
        a = np.zeros(100)
        for i in range(100):
            if i <= 10:
                args.kmean = (i+1) 
            else:
                args.kmean = 15*i
            a[i] = args.kmean
            print("Kmeans: ",args.kmean)
            acc[0][i], acc[1][i], acc[2][i] = run_model (data, train, test, train_part, args) 
            print("Accuracy test:",acc[0][i], "Accuracy train:", acc[1][i])

        plt.plot(a,1-acc[0]) 
        plt.plot(a,1-acc[1]) 
        plt.ylabel('Normalized Classification Error')
        plt.xlabel('Vocabulary size')
        plt.title('Classification error varying vocabulary size')
        plt.legend(('Test','Train'),loc='upper right')
        plt.show()
        plt.plot(a,acc[2]) 
        plt.ylabel('Time (s)')
        plt.xlabel('Vocabulary size')
        plt.title('Time complexity varying vocabulary size')
        plt.show()

    else:
        acc = run_model (data, train, test, train_part, args)
        print(acc)
    
if __name__ == "__main__":
    main()