def plot_pokemon(id=random.randint(1, 700)):
'''
plots images of a pokemon given id
across all generations of games and data available
'''
path = "data/main-sprites/"
game_folders = sorted(
[x for x in os.listdir(path) if not x.startswith('.')])
image = "{n}.png".format(n=id)
plt.figure(figsize=(20, 4))
# plot pokemon from all folders
counter = 1
for folder in game_folders:
try:
img = mpimg.imread(os.path.join(path, folder, image))
plt.subplot(2, 8, counter)
plt.imshow(img)
plt.title(str(folder))
counter += 1
except FileNotFoundError:
pass
# can't forget the icon
img = mpimg.imread(os.path.join("data/icons/", image))
plt.subplot(2, 8, counter)
plt.imshow(img)
plt.title("icon")
# we'll tell you the name as well
plt.suptitle("Pokemon #" + str(id) + " " + get_pokemon(str(id)).name)
plt.show()
def predict_single(img, classifier, test_set):
''' predicts the type of the pokemon given by "img" '''
test_image = load_image(img)
result1 = classifier.predict_classes(test_image)
predicted_type = [
type for type, index in test_set.class_indices.items()
if index == result1[0]
][0]
pokemon = get_pokemon(img)
return predicted_type
def run(evaluate=True, predict=True):
classifier, classifier2 = load_models()
# tests
test = 'type1_sorted/test'
path = test + '/' + random.choice(
[x for x in os.listdir(test) if os.path.isdir(os.path.join(test, x))])
imagepath = path + '/' + random.choice(os.listdir(path))
# load test sets
test_datagen = ImageDataGenerator(rescale=1. / 255)
test_set = test_datagen.flow_from_directory(test,
target_size=(32, 32),
batch_size=32,
class_mode='categorical')
test_set2 = test_datagen.flow_from_directory('type2_sorted/test',
target_size=(32, 32),
batch_size=32,
class_mode='categorical')
if predict:
test_image = load_image(imagepath)
result1 = classifier.predict_classes(test_image)
result2 = classifier2.predict_classes(test_image)
predicted_type = [
type for type, index in test_set.class_indices.items()
if index == result1[0]
][0]
predicted_type2 = [
type for type, index in test_set2.class_indices.items()
if index == result2[0]
][0]
pokemon = get_pokemon(imagepath)
second = "" if pokemon.type2 == 'None' else " and type " + pokemon.type2
pred_sec = "" if predicted_type2 == 'None' else " and " + predicted_type2
print("The Pokemon " + pokemon.name + " has type " + pokemon.type1 +
second)
print("The predicted type is " + predicted_type + second)
if evaluate:
print("Primary Type:")
accuracy = classifier.evaluate_generator(test_set)
print("Loss: ", accuracy[0])
print("Accuracy: ", accuracy[1])
print("Secondary:")
accuracy2 = classifier2.evaluate_generator(test_set2)
print("Loss: ", accuracy2[0])
print("Accuracy: ", accuracy2[1])
def labels(window, type_labels, pokemon_id, player_counter, AI_counter,
prediction, classifier, test_set, name_txt):
typing = types('data/Pokemon-2.csv')
path = "data/type_labels/"
type = get_pokemon(pokemon_id).type1
AI_answer = (type == prediction)
type_labels[0] = (path + type + '.png', True)
for i in range(1, 4):
name = path + random.choice(os.listdir(path))
temp = [a for b in type_labels for a in b]
while str(name).endswith('.DS_Store') or name in temp:
name = path + random.choice(os.listdir(path))
type_labels[i] = (name, False)
for i in range(4):
pilImage = Image.open(type_labels[i][0])
image = ImageTk.PhotoImage(pilImage)
type_labels[i] = (image, type_labels[i][1])
random.shuffle(type_labels)
btn1 = Button(window,
image=type_labels[0][0],
command=lambda: clicked(window, type_labels, type_labels[0][
1], player_counter, AI_counter, AI_answer, classifier,
test_set, name_txt))
btn2 = Button(window,
image=type_labels[1][0],
command=lambda: clicked(window, type_labels, type_labels[1][
1], player_counter, AI_counter, AI_answer, classifier,
test_set, name_txt))
btn3 = Button(window,
image=type_labels[2][0],
command=lambda: clicked(window, type_labels, type_labels[2][
1], player_counter, AI_counter, AI_answer, classifier,
test_set, name_txt))
btn4 = Button(window,
image=type_labels[3][0],
command=lambda: clicked(window, type_labels, type_labels[3][
1], player_counter, AI_counter, AI_answer, classifier,
test_set, name_txt))
btn1.place(x=100, y=200, anchor="center")
btn2.place(x=100, y=250, anchor="center")
btn3.place(x=200, y=200, anchor="center")
btn4.place(x=200, y=250, anchor="center")
def next_pokemon(window, classifier, test_set, name_txt):
path = "data/main-sprites/"
game_vers = random.choice(os.listdir(path))
while str(game_vers).endswith('.DS_Store'):
game_vers = random.choice(os.listdir(path))
img = random.choice(os.listdir(path + game_vers))
while not str(img).endswith('png'):
img = random.choice(os.listdir(path + game_vers))
file_name = path + game_vers + '/' + img
pilImage = Image.open(file_name)
pilImage = pilImage.resize((100, 100), Image.ANTIALIAS)
image = ImageTk.PhotoImage(pilImage)
p = Label(window, image=image)
p.photo = image
p.place(x=150, y=130, anchor="center")
name_txt.set(get_pokemon(img).name)
prediction = predict_single(file_name, classifier, test_set)
return img, prediction
def plot_game(game, pokemons = random.sample(range(1, 152), 10)):
'''
plots list of pokemons in a given game or the icons folder
'''
path = "data/main-sprites/"
full_path = os.path.join(path, game) if game is not "icons" else "data/icons"
batch = len(pokemons)
plt.figure(figsize = (20, 4))
# plot the pokemon from pokemons
counter = 1
for id in pokemons:
try:
image = "{n}.png".format(n = id)
img = mpimg.imread(os.path.join(full_path, image))
plt.subplot(1, batch, counter)
plt.imshow(img)
plt.title(get_pokemon(str(id)).name)
counter += 1
except FileNotFoundError:
pass
plt.suptitle("Pokemons from " + str(game))
plt.show()
def run(evaluate=True, predict=True):
'''
function to perform prediction using models
evaluate determines whether to evaluate model through
the Keras evaluation function, computing precision/recall statistics
predict will sample a random pokemon image to predict
'''
classifier, classifier2 = load_models()
# tests
test = 'type1_sorted/test'
path = test + '/' + random.choice(
[x for x in os.listdir(test) if os.path.isdir(os.path.join(test, x))])
imagepath = path + '/' + random.choice(os.listdir(path))
# load test sets
test_datagen = ImageDataGenerator(rescale=1. / 255)
test_set = test_datagen.flow_from_directory(test,
target_size=(32, 32),
class_mode='categorical')
test_set2 = test_datagen.flow_from_directory('type2_sorted/test',
target_size=(32, 32),
batch_size=32,
class_mode='categorical')
if predict:
# predict random pokemon
test_image = load_image(imagepath)
result1 = classifier.predict_classes(test_image)
result2 = classifier2.predict_classes(test_image)
predicted_type = [
type for type, index in test_set.class_indices.items()
if index == result1[0]
][0]
predicted_type2 = [
type for type, index in test_set2.class_indices.items()
if index == result2[0]
][0]
pokemon = get_pokemon(imagepath)
second = "" if pokemon.type2 == 'None' else " and type " + pokemon.type2
pred_sec = "" if predicted_type2 == 'None' else " and " + predicted_type2
print("The Pokemon " + pokemon.name + " has type " + pokemon.type1 +
second)
print("The predicted type is " + predicted_type + pred_sec)
plot_type(imagepath,
predicted=True,
pred_types=[predicted_type, predicted_type2])
if evaluate:
# evaluates models
print("Primary Type:")
accuracy = classifier.evaluate_generator(test_set)
v_t, v_p = stats(test_set.class_indices, classifier)
report = classification_report(v_t,
v_p,
target_names=list(
test_set.class_indices.keys()))
print(report)
print("Accuracy: ", accuracy_score(v_t, v_p))
print("\n ------------------------------- \n")
print("Evaluation Statistics")
print("Loss: ", accuracy[0])
print("Accuracy: ", accuracy[1])
print("\n ------------------------------- \n")
print("Secondary Type:")
accuracy2 = classifier2.evaluate_generator(test_set2)
v_t, v_p = stats(test_set2.class_indices, classifier2, primary=False)
report = classification_report(v_t,
v_p,
target_names=list(
test_set2.class_indices.keys()))
print(report)
print("Accuracy: ", accuracy_score(v_t, v_p))
print("\n ------------------------------- \n")
print("Evaluation Statistics")
print("Loss: ", accuracy2[0])
print("Accuracy: ", accuracy2[1])
return classifier, test_set
def plot_type(imgpath = "", predicted = False, pred_types = []):
'''
plots the true type of the pokemon with its image
if predicted, then will plot the predicted types as well
'''
pkmn = get_pokemon(imgpath)
dual_type = False if pkmn.type2 is "None" else True
grid_rows = 3 if predicted else 2
grid_cols = 2 if dual_type else 1
figsize = (4, 4)
width_ratio = (1, 1) if dual_type else (1, )
height_ratio = (12, 1, 1) if predicted else (12, 1)
if dual_type:
plt.figure(figsize = figsize)
gs = gridspec.GridSpec(grid_rows, grid_cols, height_ratios = height_ratio, width_ratios = width_ratio)
gs.update(wspace= 0.1, hspace= 0.5)
ax0 = plt.subplot(gs[0, :])
image = mpimg.imread(imgpath)
img = ax0.imshow(image)
plt.axis('off')
ax1 = plt.subplot(gs[1, 0])
typepath = "data/type_labels/{n}.png".format(n = pkmn.type1)
typeicon = mpimg.imread(typepath)
ax1.imshow(typeicon)
plt.axis('off')
ax1 = plt.subplot(gs[1, 1])
type2path = "data/type_labels/{n}.png".format(n = pkmn.type2)
type2icon = mpimg.imread(type2path)
ax1.imshow(type2icon)
plt.axis('off')
else:
plt.figure(figsize = figsize)
gs = gridspec.GridSpec(grid_rows, grid_cols, height_ratios = height_ratio, width_ratios = width_ratio)
gs.update(wspace= 0.25, hspace= 0.5)
ax0 = plt.subplot(gs[0])
image = mpimg.imread(imgpath)
img = ax0.imshow(image)
plt.axis('off')
ax1 = plt.subplot(gs[1])
typepath = "data/type_labels/{n}.png".format(n = pkmn.type1)
typeicon = mpimg.imread(typepath)
ax1.imshow(typeicon)
plt.axis('off')
if predicted:
if dual_type:
ax3 = plt.subplot(gs[2, 0])
plt.text(-80, 12, "Predicted:")
typepath = "data/type_labels/{n}.png".format(n = pred_types[0])
typeicon = mpimg.imread(typepath)
ax3.imshow(typeicon)
plt.axis('off')
ax4 = plt.subplot(gs[2, 1])
try:
typepath = "data/type_labels/{n}.png".format(n = pred_types[1])
typeicon = mpimg.imread(typepath)
ax4.imshow(typeicon)
except FileNotFoundError:
pass
plt.axis('off')
else:
ax3 = plt.subplot(gs[2, :])
plt.text(-80, 12, "Predicted:")
typepath = "data/type_labels/{n}.png".format(n = pred_types[0])
typeicon = mpimg.imread(typepath)
ax3.imshow(typeicon)
plt.axis('off')
plt.suptitle(pkmn.name)
plt.show()