Add scripts

src/scripts/mnist.py

@@ -0,0 +1,32 @@
+#!/usr/bin/python3
+IMAGES_MAGIC_NUMBER = 2051
+LABELS_MAGIC_NUMBER = 2049
+
+
+def write_images_to_binary(images, filepath):
+    byteorder = "big"
+
+    bytes_ = IMAGES_MAGIC_NUMBER.to_bytes(4, byteorder=byteorder)
+    bytes_ += (len(images)).to_bytes(4, byteorder=byteorder)
+    bytes_ += len(images[0][0]).to_bytes(4, byteorder=byteorder)
+    bytes_ += len(images[0]).to_bytes(4, byteorder=byteorder)
+
+    for image in images:
+        for row in image:
+            for nb in row:
+                bytes_ += int(nb).to_bytes(1, byteorder=byteorder)
+
+    with open(filepath, "wb") as f:
+        f.write(bytes_)
+
+
+def write_labels_to_binary(labels, filepath):
+    byteorder = "big"
+
+    bytes_ = LABELS_MAGIC_NUMBER.to_bytes(4, byteorder=byteorder)
+    bytes_ += (len(labels)).to_bytes(4, byteorder=byteorder)
+    for label in labels:
+        bytes_ += int(label).to_bytes(1, byteorder=byteorder)
+
+    with open(filepath, "wb") as f:
+        f.write(bytes_)

src/scripts/visualisation.py

@@ -0,0 +1,119 @@
+#!/usr/bin/python3
+"""
+Ensemble de fonctions permettant de visualiser
+les différentes données disponibles dans le réseau de neurones
+"""
+import json
+import math
+import os
+
+import png
+from matplotlib import pyplot as plt
+
+IMAGE_WIDTH = 28
+IMAGE_HEIGHT = 28
+
+def image_from_file(filepath, dest="./images/"):
+    """
+    Enregistre un ensemble d'images au format PNG
+    à partir d'un fichier texte comprenant une liste d'images
+    chaque image étant un tableau de poids entre 0 et 255
+    """
+    os.makedirs(dest, exist_ok=True)
+    with open(filepath, "r", encoding="utf8") as fp:
+        data = json.load(fp)
+
+    nb_elem = len(data)
+    for i in range(nb_elem):
+        png.from_array(data[i], 'L').save(os.path.join(dest, f"{i}.png"))
+
+
+def image_from_list(filepath, dest="data.png", exp=False):
+    """
+    Enregistre une liste de poids sous forme d'une image
+    exp sert à spécifier si il faut passer à une forme exponentielle
+    afin de mieux distinguer les points prédominants.
+    """
+    with open(filepath, "r", encoding="utf8") as fp:
+        data = json.load(fp)
+
+    mini = min(data)
+    data = [i-mini for i in data] # Set min to 0
+
+    maxi = max(data)
+    if exp:
+        ratio = 255/math.exp(maxi)
+        data = [int(math.exp(i)*ratio) for i in data]
+    else:
+        ratio = 255/maxi
+        data = [int(i*ratio) for i in data]
+
+    new_data = [[0 for i in range(IMAGE_WIDTH)] for j in range(IMAGE_HEIGHT)]
+
+    for i in range(IMAGE_WIDTH):
+        for j in range(IMAGE_HEIGHT):
+            new_data[i][j] = data[i*IMAGE_HEIGHT+j]
+
+    png.from_array(new_data, 'L').save(dest)
+
+
+def graph_from_test_reseau(erreurs, reussites):
+    """
+    Affiche un graphique à partir d'un fichier contenant les
+    réussites et d'un autre contenant les erreurs (sortie brutes de out/main)
+    """
+    with open(erreurs, "r", encoding="utf8") as f:
+        data = f.read()
+
+    data = data.split("--- Image")[1:]
+    data = [i.split("\n")[:IMAGE_HEIGHT] for i in data]
+    labels = []
+    for i in data:
+        labels.append((int(i[0].split(",")[1][1]), int(i[0][-5])))
+
+    data = [[float(j[IMAGE_HEIGHT+5:]) for j in i if j[IMAGE_HEIGHT+5:] != ''] for i in data]
+
+    x = []
+    y = []
+
+    for i, label in enumerate(labels):
+        x.append(data[i][label[0]])
+        y.append(data[i][label[1]])
+
+    plt.plot(x, y, "+r")
+
+    with open(reussites, "r", encoding="utf8") as f:
+        data = f.read()
+
+    data = data.split("--- Image")[1:]
+    data = [i.split("\n")[:IMAGE_HEIGHT] for i in data]
+    labels = []
+    for i in data:
+        labels.append((int(i[0].split(",")[1][1]), int(i[0][-5])))
+
+    data = [[float(j[IMAGE_HEIGHT+5:]) for j in i if j[IMAGE_HEIGHT+5:] != ''] for i in data]
+
+    x = []
+    y = []
+
+    for i, label in enumerate(labels):
+        x.append(data[i][label[0]])
+        y.append(data[i][label[1]])
+
+    plt.plot(x, y, "+b")
+    plt.xlabel("Réel")
+    plt.ylabel("Prévision")
+    plt.legend()
+    plt.show()
+
+
+def images_neurons(neurons, dest="neurons", exp=False):
+    """
+    Appelle le programme C ainsi que la fonction image_from_list
+    Afin de créer un ensemble d'image visualisant les poids
+    """
+    os.makedirs(dest, exist_ok=True)
+    for i in neurons:
+        os.system(f"./make.sh utils print-poids-neurone --reseau \
+            .cache/reseau.bin --neurone {i} > .cache/poids.txt")
+        image_from_list(".cache/poids.txt", os.path.join(dest, f"{i}.png"), exp=exp)