Butterfly Recognition¶
The datset for this notebook is taken from - here.
Train, Test. Validation data set for 50 butterfly species. All images are 224 X 224 X 3 in jpg format . Train set consists of 4955 images partitioned into 50 sub directories one for each species. Test set consists of 250 images partitioned into 50 sub directories with 5 test images per species. Valid set consists of 250 images partitioned into 50 sub directories with 5 validation images per species. 1 CSV file is included with 3 columns filepaths, labels and dataset. filepaths is the relative path to an image. labels is the string name of the species associate with the image file. dataset species the directory (train, test, valid) that contains the image file.
The model classifies for the given classes - 'adonis','american snoot','an 88','banded peacock','beckers white','black hairstreak', 'cabbage white','chestnut','clodius parnassian','clouded sulphur','copper tail','crecent','crimson patch','eastern coma','gold banded', 'great eggfly','grey hairstreak','indra swallow','julia','large marble','malachite','mangrove skipper','metalmark','monarch','morning cloak', 'orange oakleaf','orange tip','orchard swallow','painted lady','paper kite','peacock','pine white','pipevine swallow','purple hairstreak','question mark', 'red admiral','red spotted purple','scarce swallow','silver spot skipper','sixspot burnet','skipper','sootywing','southern dogface','straited queen', 'two barred flasher','ulyses','viceroy','wood satyr','yellow swallow tail','zebra long wing'.
Getting our data using Kaggle API¶
In [ ]:
Copied!
! mkdir ~/.kaggle
! cp kaggle.json ~/.kaggle/
! chmod 600 ~/.kaggle/kaggle.json
! kaggle datasets download -d gpiosenka/butterfly-images40-species
! mkdir ~/.kaggle
! cp kaggle.json ~/.kaggle/
! chmod 600 ~/.kaggle/kaggle.json
! kaggle datasets download -d gpiosenka/butterfly-images40-species
Downloading butterfly-images40-species.zip to /content 99% 835M/842M [00:07<00:00, 122MB/s] 100% 842M/842M [00:07<00:00, 118MB/s]
In [ ]:
Copied!
# To view which GPU are we running
! nvidia-smi -L
# To view which GPU are we running
! nvidia-smi -L
GPU 0: Tesla K80 (UUID: GPU-4fea7eb5-210a-fca4-101e-ed85517559cf)
In [ ]:
Copied!
# downloading helper_functions.py
! wget https://raw.githubusercontent.com/Hrushi11/Dogs_VS_Cats/main/helper_functions.py
# downloading helper_functions.py
! wget https://raw.githubusercontent.com/Hrushi11/Dogs_VS_Cats/main/helper_functions.py
--2021-08-02 16:57:39-- https://raw.githubusercontent.com/Hrushi11/Dogs_VS_Cats/main/helper_functions.py Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.110.133, 185.199.109.133, 185.199.111.133, ... Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.110.133|:443... connected. HTTP request sent, awaiting response... 200 OK Length: 10139 (9.9K) [text/plain] Saving to: ‘helper_functions.py’ helper_functions.py 100%[===================>] 9.90K --.-KB/s in 0s 2021-08-02 16:57:39 (82.5 MB/s) - ‘helper_functions.py’ saved [10139/10139]
Importing Dependancies¶
In [ ]:
Copied!
# Importing dependancies
import os
import random
import numpy as np
import pandas as pd
import tensorflow as tf
import matplotlib.pyplot as plt
from sklearn.metrics import classification_report
from tensorflow.keras.layers.experimental import preprocessing
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from helper_functions import unzip_data, walk_through_dir, plot_loss_curves, make_confusion_matrix, compare_historys
# Importing dependancies
import os
import random
import numpy as np
import pandas as pd
import tensorflow as tf
import matplotlib.pyplot as plt
from sklearn.metrics import classification_report
from tensorflow.keras.layers.experimental import preprocessing
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from helper_functions import unzip_data, walk_through_dir, plot_loss_curves, make_confusion_matrix, compare_historys
In [ ]:
Copied!
# Unzipping our data
unzip_data("/content/butterfly-images40-species.zip")
# Unzipping our data
unzip_data("/content/butterfly-images40-species.zip")
In [ ]:
Copied!
# deleting the zip file
!rm butterfly-images40-species.zip
# deleting the zip file
!rm butterfly-images40-species.zip
In [ ]:
Copied!
# deleting the extra folder
!rm -rf butterflies_rev2
# deleting the extra folder
!rm -rf butterflies_rev2
In [ ]:
Copied!
# A look through all the directories in the dataset
walk_through_dir("/content/butterflies")
# A look through all the directories in the dataset
walk_through_dir("/content/butterflies")
There are 3 directories and 1 images in '/content/butterflies'. There are 50 directories and 0 images in '/content/butterflies/valid'. There are 0 directories and 5 images in '/content/butterflies/valid/straited queen'. There are 0 directories and 5 images in '/content/butterflies/valid/paper kite'. There are 0 directories and 5 images in '/content/butterflies/valid/beckers white'. There are 0 directories and 5 images in '/content/butterflies/valid/sootywing'. There are 0 directories and 5 images in '/content/butterflies/valid/clouded sulphur'. There are 0 directories and 5 images in '/content/butterflies/valid/monarch'. There are 0 directories and 5 images in '/content/butterflies/valid/two barred flasher'. There are 0 directories and 5 images in '/content/butterflies/valid/indra swallow'. There are 0 directories and 5 images in '/content/butterflies/valid/copper tail'. There are 0 directories and 5 images in '/content/butterflies/valid/wood satyr'. There are 0 directories and 5 images in '/content/butterflies/valid/mangrove skipper'. There are 0 directories and 5 images in '/content/butterflies/valid/crecent'. There are 0 directories and 5 images in '/content/butterflies/valid/an 88'. There are 0 directories and 5 images in '/content/butterflies/valid/ulyses'. There are 0 directories and 5 images in '/content/butterflies/valid/american snoot'. There are 0 directories and 5 images in '/content/butterflies/valid/red admiral'. There are 0 directories and 5 images in '/content/butterflies/valid/scarce swallow'. There are 0 directories and 5 images in '/content/butterflies/valid/pipevine swallow'. There are 0 directories and 5 images in '/content/butterflies/valid/painted lady'. There are 0 directories and 5 images in '/content/butterflies/valid/skipper'. There are 0 directories and 5 images in '/content/butterflies/valid/red spotted purple'. There are 0 directories and 5 images in '/content/butterflies/valid/orchard swallow'. There are 0 directories and 5 images in '/content/butterflies/valid/chestnut'. There are 0 directories and 5 images in '/content/butterflies/valid/viceroy'. There are 0 directories and 5 images in '/content/butterflies/valid/yellow swallow tail'. There are 0 directories and 5 images in '/content/butterflies/valid/crimson patch'. There are 0 directories and 5 images in '/content/butterflies/valid/orange oakleaf'. There are 0 directories and 5 images in '/content/butterflies/valid/large marble'. There are 0 directories and 5 images in '/content/butterflies/valid/question mark'. There are 0 directories and 5 images in '/content/butterflies/valid/banded peacock'. There are 0 directories and 5 images in '/content/butterflies/valid/silver spot skipper'. There are 0 directories and 5 images in '/content/butterflies/valid/sixspot burnet'. There are 0 directories and 5 images in '/content/butterflies/valid/eastern coma'. There are 0 directories and 5 images in '/content/butterflies/valid/southern dogface'. There are 0 directories and 5 images in '/content/butterflies/valid/clodius parnassian'. There are 0 directories and 5 images in '/content/butterflies/valid/malachite'. There are 0 directories and 5 images in '/content/butterflies/valid/morning cloak'. There are 0 directories and 5 images in '/content/butterflies/valid/peacock'. There are 0 directories and 5 images in '/content/butterflies/valid/grey hairstreak'. There are 0 directories and 5 images in '/content/butterflies/valid/purple hairstreak'. There are 0 directories and 5 images in '/content/butterflies/valid/gold banded'. There are 0 directories and 5 images in '/content/butterflies/valid/cabbage white'. There are 0 directories and 5 images in '/content/butterflies/valid/zebra long wing'. There are 0 directories and 5 images in '/content/butterflies/valid/orange tip'. There are 0 directories and 5 images in '/content/butterflies/valid/pine white'. There are 0 directories and 5 images in '/content/butterflies/valid/metalmark'. There are 0 directories and 5 images in '/content/butterflies/valid/black hairstreak'. There are 0 directories and 5 images in '/content/butterflies/valid/julia'. There are 0 directories and 5 images in '/content/butterflies/valid/adonis'. There are 0 directories and 5 images in '/content/butterflies/valid/great eggfly'. There are 50 directories and 0 images in '/content/butterflies/train'. There are 0 directories and 96 images in '/content/butterflies/train/straited queen'. There are 0 directories and 92 images in '/content/butterflies/train/paper kite'. There are 0 directories and 116 images in '/content/butterflies/train/beckers white'. There are 0 directories and 128 images in '/content/butterflies/train/sootywing'. There are 0 directories and 95 images in '/content/butterflies/train/clouded sulphur'. There are 0 directories and 80 images in '/content/butterflies/train/monarch'. There are 0 directories and 82 images in '/content/butterflies/train/two barred flasher'. There are 0 directories and 115 images in '/content/butterflies/train/indra swallow'. There are 0 directories and 134 images in '/content/butterflies/train/copper tail'. There are 0 directories and 102 images in '/content/butterflies/train/wood satyr'. There are 0 directories and 90 images in '/content/butterflies/train/mangrove skipper'. There are 0 directories and 66 images in '/content/butterflies/train/crecent'. There are 0 directories and 92 images in '/content/butterflies/train/an 88'. There are 0 directories and 72 images in '/content/butterflies/train/ulyses'. There are 0 directories and 105 images in '/content/butterflies/train/american snoot'. There are 0 directories and 91 images in '/content/butterflies/train/red admiral'. There are 0 directories and 139 images in '/content/butterflies/train/scarce swallow'. There are 0 directories and 94 images in '/content/butterflies/train/pipevine swallow'. There are 0 directories and 112 images in '/content/butterflies/train/painted lady'. There are 0 directories and 136 images in '/content/butterflies/train/skipper'. There are 0 directories and 95 images in '/content/butterflies/train/red spotted purple'. There are 0 directories and 81 images in '/content/butterflies/train/orchard swallow'. There are 0 directories and 90 images in '/content/butterflies/train/chestnut'. There are 0 directories and 88 images in '/content/butterflies/train/viceroy'. There are 0 directories and 107 images in '/content/butterflies/train/yellow swallow tail'. There are 0 directories and 103 images in '/content/butterflies/train/crimson patch'. There are 0 directories and 100 images in '/content/butterflies/train/orange oakleaf'. There are 0 directories and 79 images in '/content/butterflies/train/large marble'. There are 0 directories and 110 images in '/content/butterflies/train/question mark'. There are 0 directories and 91 images in '/content/butterflies/train/banded peacock'. There are 0 directories and 94 images in '/content/butterflies/train/silver spot skipper'. There are 0 directories and 101 images in '/content/butterflies/train/sixspot burnet'. There are 0 directories and 96 images in '/content/butterflies/train/eastern coma'. There are 0 directories and 94 images in '/content/butterflies/train/southern dogface'. There are 0 directories and 77 images in '/content/butterflies/train/clodius parnassian'. There are 0 directories and 104 images in '/content/butterflies/train/malachite'. There are 0 directories and 90 images in '/content/butterflies/train/morning cloak'. There are 0 directories and 97 images in '/content/butterflies/train/peacock'. There are 0 directories and 98 images in '/content/butterflies/train/grey hairstreak'. There are 0 directories and 113 images in '/content/butterflies/train/purple hairstreak'. There are 0 directories and 104 images in '/content/butterflies/train/gold banded'. There are 0 directories and 92 images in '/content/butterflies/train/cabbage white'. There are 0 directories and 108 images in '/content/butterflies/train/zebra long wing'. There are 0 directories and 90 images in '/content/butterflies/train/orange tip'. There are 0 directories and 96 images in '/content/butterflies/train/pine white'. There are 0 directories and 108 images in '/content/butterflies/train/metalmark'. There are 0 directories and 90 images in '/content/butterflies/train/black hairstreak'. There are 0 directories and 115 images in '/content/butterflies/train/julia'. There are 0 directories and 96 images in '/content/butterflies/train/adonis'. There are 0 directories and 111 images in '/content/butterflies/train/great eggfly'. There are 50 directories and 0 images in '/content/butterflies/test'. There are 0 directories and 5 images in '/content/butterflies/test/straited queen'. There are 0 directories and 5 images in '/content/butterflies/test/paper kite'. There are 0 directories and 5 images in '/content/butterflies/test/beckers white'. There are 0 directories and 5 images in '/content/butterflies/test/sootywing'. There are 0 directories and 5 images in '/content/butterflies/test/clouded sulphur'. There are 0 directories and 5 images in '/content/butterflies/test/monarch'. There are 0 directories and 5 images in '/content/butterflies/test/two barred flasher'. There are 0 directories and 5 images in '/content/butterflies/test/indra swallow'. There are 0 directories and 5 images in '/content/butterflies/test/copper tail'. There are 0 directories and 5 images in '/content/butterflies/test/wood satyr'. There are 0 directories and 5 images in '/content/butterflies/test/mangrove skipper'. There are 0 directories and 5 images in '/content/butterflies/test/crecent'. There are 0 directories and 5 images in '/content/butterflies/test/an 88'. There are 0 directories and 5 images in '/content/butterflies/test/ulyses'. There are 0 directories and 5 images in '/content/butterflies/test/american snoot'. There are 0 directories and 5 images in '/content/butterflies/test/red admiral'. There are 0 directories and 5 images in '/content/butterflies/test/scarce swallow'. There are 0 directories and 5 images in '/content/butterflies/test/pipevine swallow'. There are 0 directories and 5 images in '/content/butterflies/test/painted lady'. There are 0 directories and 5 images in '/content/butterflies/test/skipper'. There are 0 directories and 5 images in '/content/butterflies/test/red spotted purple'. There are 0 directories and 5 images in '/content/butterflies/test/orchard swallow'. There are 0 directories and 5 images in '/content/butterflies/test/chestnut'. There are 0 directories and 5 images in '/content/butterflies/test/viceroy'. There are 0 directories and 5 images in '/content/butterflies/test/yellow swallow tail'. There are 0 directories and 5 images in '/content/butterflies/test/crimson patch'. There are 0 directories and 5 images in '/content/butterflies/test/orange oakleaf'. There are 0 directories and 5 images in '/content/butterflies/test/large marble'. There are 0 directories and 5 images in '/content/butterflies/test/question mark'. There are 0 directories and 5 images in '/content/butterflies/test/banded peacock'. There are 0 directories and 5 images in '/content/butterflies/test/silver spot skipper'. There are 0 directories and 5 images in '/content/butterflies/test/sixspot burnet'. There are 0 directories and 5 images in '/content/butterflies/test/eastern coma'. There are 0 directories and 5 images in '/content/butterflies/test/southern dogface'. There are 0 directories and 5 images in '/content/butterflies/test/clodius parnassian'. There are 0 directories and 5 images in '/content/butterflies/test/malachite'. There are 0 directories and 5 images in '/content/butterflies/test/morning cloak'. There are 0 directories and 5 images in '/content/butterflies/test/peacock'. There are 0 directories and 5 images in '/content/butterflies/test/grey hairstreak'. There are 0 directories and 5 images in '/content/butterflies/test/purple hairstreak'. There are 0 directories and 5 images in '/content/butterflies/test/gold banded'. There are 0 directories and 5 images in '/content/butterflies/test/cabbage white'. There are 0 directories and 5 images in '/content/butterflies/test/zebra long wing'. There are 0 directories and 5 images in '/content/butterflies/test/orange tip'. There are 0 directories and 5 images in '/content/butterflies/test/pine white'. There are 0 directories and 5 images in '/content/butterflies/test/metalmark'. There are 0 directories and 5 images in '/content/butterflies/test/black hairstreak'. There are 0 directories and 5 images in '/content/butterflies/test/julia'. There are 0 directories and 5 images in '/content/butterflies/test/adonis'. There are 0 directories and 5 images in '/content/butterflies/test/great eggfly'.
In [ ]:
Copied!
labels = os.listdir("/content/butterflies/train")
len(labels)
labels = os.listdir("/content/butterflies/train")
len(labels)
Out[ ]:
50
Getting our data ready¶
In [ ]:
Copied!
# creating a list of all the training file paths
train_file_paths = []
for elem in labels:
dir_path = "/content/butterflies/train/" + elem
paths = os.listdir(dir_path)
for path in paths:
path = dir_path + "/" + path
train_file_paths.append(path)
# creating a list of all the training file paths
train_file_paths = []
for elem in labels:
dir_path = "/content/butterflies/train/" + elem
paths = os.listdir(dir_path)
for path in paths:
path = dir_path + "/" + path
train_file_paths.append(path)
In [ ]:
Copied!
len(train_file_paths)
len(train_file_paths)
Out[ ]:
4955
In [ ]:
Copied!
# randomly checking a part of the list
train_file_paths[100:110]
# randomly checking a part of the list
train_file_paths[100:110]
Out[ ]:
['/content/butterflies/train/paper kite/67.jpg', '/content/butterflies/train/paper kite/43.jpg', '/content/butterflies/train/paper kite/10.jpg', '/content/butterflies/train/paper kite/24.jpg', '/content/butterflies/train/paper kite/03.jpg', '/content/butterflies/train/paper kite/27.jpg', '/content/butterflies/train/paper kite/28.jpg', '/content/butterflies/train/paper kite/57.jpg', '/content/butterflies/train/paper kite/23.jpg', '/content/butterflies/train/paper kite/55.jpg']
In [ ]:
Copied!
# To read and preprocess the image
def load_and_preprocess(filepath):
"""
Takes a file path and preprocess images into 224 x 3
"""
img_file = tf.io.read_file(filepath)
img = tf.image.decode_image(img_file)
img = tf.image.resize(img, (224, 224))
return img
# To read and preprocess the image
def load_and_preprocess(filepath):
"""
Takes a file path and preprocess images into 224 x 3
"""
img_file = tf.io.read_file(filepath)
img = tf.image.decode_image(img_file)
img = tf.image.resize(img, (224, 224))
return img
In [ ]:
Copied!
# Visualizing random image
random_file = random.choice(train_file_paths)
img = load_and_preprocess(random_file)
plt.axis(False)
plt.title(random_file.split("/")[4], color="green")
plt.imshow(img / 255.);
# Visualizing random image
random_file = random.choice(train_file_paths)
img = load_and_preprocess(random_file)
plt.axis(False)
plt.title(random_file.split("/")[4], color="green")
plt.imshow(img / 255.);
Visualizing multiple random images¶
In [ ]:
Copied!
# Visualizing multiple random images
plt.figure(figsize=(17, 12))
for i in range(9):
# Random selecting an image from the train dataset
random_file = random.choice(train_file_paths)
img = load_and_preprocess(random_file)
# Plotting the random images
plt.subplot(3, 3, i+1)
plt.title(random_file.split("/")[4], color="green")
plt.axis(False)
plt.imshow(img / 255.)
# Visualizing multiple random images
plt.figure(figsize=(17, 12))
for i in range(9):
# Random selecting an image from the train dataset
random_file = random.choice(train_file_paths)
img = load_and_preprocess(random_file)
# Plotting the random images
plt.subplot(3, 3, i+1)
plt.title(random_file.split("/")[4], color="green")
plt.axis(False)
plt.imshow(img / 255.)
Data Augmentation layer¶
In [ ]:
Copied!
# Creating data augmentation layer
data_augmentation = tf.keras.Sequential([
preprocessing.RandomFlip("horizontal"),
preprocessing.RandomRotation(0.2),
preprocessing.RandomZoom(0.2),
preprocessing.RandomHeight(0.2),
preprocessing.RandomWidth(0.2)
], name="data_augmentation_layer")
# Creating data augmentation layer
data_augmentation = tf.keras.Sequential([
preprocessing.RandomFlip("horizontal"),
preprocessing.RandomRotation(0.2),
preprocessing.RandomZoom(0.2),
preprocessing.RandomHeight(0.2),
preprocessing.RandomWidth(0.2)
], name="data_augmentation_layer")
In [ ]:
Copied!
# Visualizing the augmented image
random_file = random.choice(train_file_paths)
img = load_and_preprocess(random_file)
aug_img = data_augmentation(tf.expand_dims(img, axis=0))
plt.axis(False)
plt.imshow(aug_img[0] / 255.)
# Visualizing the augmented image
random_file = random.choice(train_file_paths)
img = load_and_preprocess(random_file)
aug_img = data_augmentation(tf.expand_dims(img, axis=0))
plt.axis(False)
plt.imshow(aug_img[0] / 255.)
Out[ ]:
<matplotlib.image.AxesImage at 0x7f45c3766f10>
Visualizing a single augmented image¶
In [ ]:
Copied!
# Visualizing the augmentated image with the original image
random_file = random.choice(train_file_paths)
img = load_and_preprocess(random_file)
aug_img = data_augmentation(tf.expand_dims(img, axis=0))
aug_img = tf.image.resize(aug_img[0], (224, 224))
label = random_file.split("/")[4]
plt.figure(figsize=(10, 7))
# Plot for original image
plt.subplot(1, 2, 1)
plt.title(f"Original Image: {label}", color = "green")
plt.axis(False)
plt.imshow(img / 255.)
# Plot for augmented image
plt.subplot(1, 2, 2)
plt.title(f"Augmented Image: {label}", color = "blue")
plt.axis(False)
plt.imshow(aug_img / 255.);
# Visualizing the augmentated image with the original image
random_file = random.choice(train_file_paths)
img = load_and_preprocess(random_file)
aug_img = data_augmentation(tf.expand_dims(img, axis=0))
aug_img = tf.image.resize(aug_img[0], (224, 224))
label = random_file.split("/")[4]
plt.figure(figsize=(10, 7))
# Plot for original image
plt.subplot(1, 2, 1)
plt.title(f"Original Image: {label}", color = "green")
plt.axis(False)
plt.imshow(img / 255.)
# Plot for augmented image
plt.subplot(1, 2, 2)
plt.title(f"Augmented Image: {label}", color = "blue")
plt.axis(False)
plt.imshow(aug_img / 255.);
Visualizing multiple augmented images¶
In [ ]:
Copied!
# Visualizing multiple augmentated images with original images
plt.figure(figsize=(17, 12))
for i in range(1, 12, 2):
# generating random file path & aug img
random_file = random.choice(train_file_paths)
img = load_and_preprocess(random_file)
aug_img = data_augmentation(tf.expand_dims(img, axis=0))
aug_img = tf.image.resize(aug_img[0], (224, 224))
label = random_file.split("/")[4]
# plotting the original image
plt.subplot(3, 4, i)
plt.title(f"Original Image: {label}", color = "green")
plt.axis(False)
plt.imshow(img / 255.)
# plotting the augmented image
plt.subplot(3, 4, i+1)
plt.title(f"Augmented Image: {label}", color = "blue")
plt.axis(False)
plt.imshow(aug_img / 255.);
# Visualizing multiple augmentated images with original images
plt.figure(figsize=(17, 12))
for i in range(1, 12, 2):
# generating random file path & aug img
random_file = random.choice(train_file_paths)
img = load_and_preprocess(random_file)
aug_img = data_augmentation(tf.expand_dims(img, axis=0))
aug_img = tf.image.resize(aug_img[0], (224, 224))
label = random_file.split("/")[4]
# plotting the original image
plt.subplot(3, 4, i)
plt.title(f"Original Image: {label}", color = "green")
plt.axis(False)
plt.imshow(img / 255.)
# plotting the augmented image
plt.subplot(3, 4, i+1)
plt.title(f"Augmented Image: {label}", color = "blue")
plt.axis(False)
plt.imshow(aug_img / 255.);
Data Generators¶
In [ ]:
Copied!
train_data_gen = ImageDataGenerator()
valid_data_gen = ImageDataGenerator()
test_data_gen = ImageDataGenerator()
train_data_gen = ImageDataGenerator()
valid_data_gen = ImageDataGenerator()
test_data_gen = ImageDataGenerator()
In [ ]:
Copied!
train_data = train_data_gen.flow_from_directory("/content/butterflies/train",
target_size=(224, 224),
seed=42)
valid_data = valid_data_gen.flow_from_directory("/content/butterflies/valid",
target_size=(224, 224),
shuffle=False)
test_data = test_data_gen.flow_from_directory("/content/butterflies/test",
target_size=(224, 224),
shuffle=False)
train_data = train_data_gen.flow_from_directory("/content/butterflies/train",
target_size=(224, 224),
seed=42)
valid_data = valid_data_gen.flow_from_directory("/content/butterflies/valid",
target_size=(224, 224),
shuffle=False)
test_data = test_data_gen.flow_from_directory("/content/butterflies/test",
target_size=(224, 224),
shuffle=False)
Found 4955 images belonging to 50 classes. Found 250 images belonging to 50 classes. Found 250 images belonging to 50 classes.
In [ ]:
Copied!
# To get all the class names
class_ = train_data.class_indices
class_names = []
for key in class_.keys():
class_names.append(key)
# To get all the class names
class_ = train_data.class_indices
class_names = []
for key in class_.keys():
class_names.append(key)
In [ ]:
Copied!
class_names
class_names
Out[ ]:
['adonis', 'american snoot', 'an 88', 'banded peacock', 'beckers white', 'black hairstreak', 'cabbage white', 'chestnut', 'clodius parnassian', 'clouded sulphur', 'copper tail', 'crecent', 'crimson patch', 'eastern coma', 'gold banded', 'great eggfly', 'grey hairstreak', 'indra swallow', 'julia', 'large marble', 'malachite', 'mangrove skipper', 'metalmark', 'monarch', 'morning cloak', 'orange oakleaf', 'orange tip', 'orchard swallow', 'painted lady', 'paper kite', 'peacock', 'pine white', 'pipevine swallow', 'purple hairstreak', 'question mark', 'red admiral', 'red spotted purple', 'scarce swallow', 'silver spot skipper', 'sixspot burnet', 'skipper', 'sootywing', 'southern dogface', 'straited queen', 'two barred flasher', 'ulyses', 'viceroy', 'wood satyr', 'yellow swallow tail', 'zebra long wing']
Training our Models¶
Model 1¶
In [ ]:
Copied!
# Setting up base model
base_model = tf.keras.applications.EfficientNetB0(include_top=False)
base_model.trainable = False
# Setting input layer
inputs = tf.keras.layers.Input(shape=(224, 224, 3), name="input_layer")
x = data_augmentation(inputs)
x = base_model(x, training=False)
x = tf.keras.layers.GlobalAveragePooling2D(name="global_average_pooling")(x)
outputs = tf.keras.layers.Dense(len(class_names), activation="softmax", name="output_layer")(x)
model_1 = tf.keras.Model(inputs, outputs)
# compiling the model
model_1.compile(loss=tf.keras.losses.categorical_crossentropy,
optimizer=tf.keras.optimizers.Adam(),
metrics=["accuracy"])
# fit the model
history_1 = model_1.fit(train_data,
epochs=5,
steps_per_epoch=len(train_data),
validation_data=valid_data,
validation_steps=len(valid_data))
# Setting up base model
base_model = tf.keras.applications.EfficientNetB0(include_top=False)
base_model.trainable = False
# Setting input layer
inputs = tf.keras.layers.Input(shape=(224, 224, 3), name="input_layer")
x = data_augmentation(inputs)
x = base_model(x, training=False)
x = tf.keras.layers.GlobalAveragePooling2D(name="global_average_pooling")(x)
outputs = tf.keras.layers.Dense(len(class_names), activation="softmax", name="output_layer")(x)
model_1 = tf.keras.Model(inputs, outputs)
# compiling the model
model_1.compile(loss=tf.keras.losses.categorical_crossentropy,
optimizer=tf.keras.optimizers.Adam(),
metrics=["accuracy"])
# fit the model
history_1 = model_1.fit(train_data,
epochs=5,
steps_per_epoch=len(train_data),
validation_data=valid_data,
validation_steps=len(valid_data))
Epoch 1/5 155/155 [==============================] - 76s 447ms/step - loss: 2.0292 - accuracy: 0.6125 - val_loss: 1.2592 - val_accuracy: 0.8040 Epoch 2/5 155/155 [==============================] - 53s 344ms/step - loss: 0.7720 - accuracy: 0.8692 - val_loss: 0.7581 - val_accuracy: 0.8960 Epoch 3/5 155/155 [==============================] - 52s 335ms/step - loss: 0.5180 - accuracy: 0.9049 - val_loss: 0.5798 - val_accuracy: 0.9000 Epoch 4/5 155/155 [==============================] - 44s 285ms/step - loss: 0.4050 - accuracy: 0.9255 - val_loss: 0.4514 - val_accuracy: 0.9480 Epoch 5/5 155/155 [==============================] - 43s 275ms/step - loss: 0.3350 - accuracy: 0.9338 - val_loss: 0.4040 - val_accuracy: 0.9400
In [ ]:
Copied!
model_1.evaluate(test_data)
model_1.evaluate(test_data)
8/8 [==============================] - 1s 128ms/step - loss: 0.4121 - accuracy: 0.9240
Out[ ]:
[0.41212189197540283, 0.9240000247955322]
Loss curves for original model¶
In [ ]:
Copied!
plot_loss_curves(history_1)
plot_loss_curves(history_1)
In [ ]:
Copied!
model_1.save("/content/drive/MyDrive/Butterfly-Recognition/Model_1")
model_1.save("/content/drive/MyDrive/Butterfly-Recognition/Model_1")
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument. category=CustomMaskWarning)
INFO:tensorflow:Assets written to: /content/drive/MyDrive/Butterfly-Recognition/Model_1/assets
Fine tuning the model¶
In [ ]:
Copied!
# Unfreeze all of the layers in the base model
base_model.trainable = True
# Refreeze every layer except for the last 5
for layer in base_model.layers[:-5]:
layer.trainable = False
# Recompile model with lower learning rate
model_1.compile(loss='categorical_crossentropy',
optimizer=tf.keras.optimizers.Adam(1e-4),
metrics=['accuracy'])
# Fine-tune for 5 more epochs
fine_tune_epochs = 10
history_1_fine_tune_1 = model_1.fit(train_data,
epochs=fine_tune_epochs,
validation_data=valid_data,
validation_steps= len(valid_data),
initial_epoch=history_1.epoch[-1])
# Unfreeze all of the layers in the base model
base_model.trainable = True
# Refreeze every layer except for the last 5
for layer in base_model.layers[:-5]:
layer.trainable = False
# Recompile model with lower learning rate
model_1.compile(loss='categorical_crossentropy',
optimizer=tf.keras.optimizers.Adam(1e-4),
metrics=['accuracy'])
# Fine-tune for 5 more epochs
fine_tune_epochs = 10
history_1_fine_tune_1 = model_1.fit(train_data,
epochs=fine_tune_epochs,
validation_data=valid_data,
validation_steps= len(valid_data),
initial_epoch=history_1.epoch[-1])
Epoch 5/10 155/155 [==============================] - 53s 291ms/step - loss: 0.2435 - accuracy: 0.9413 - val_loss: 0.2148 - val_accuracy: 0.9360 Epoch 6/10 155/155 [==============================] - 39s 249ms/step - loss: 0.1956 - accuracy: 0.9457 - val_loss: 0.1746 - val_accuracy: 0.9760 Epoch 7/10 155/155 [==============================] - 41s 263ms/step - loss: 0.1662 - accuracy: 0.9528 - val_loss: 0.1651 - val_accuracy: 0.9680 Epoch 8/10 155/155 [==============================] - 38s 247ms/step - loss: 0.1537 - accuracy: 0.9590 - val_loss: 0.1452 - val_accuracy: 0.9640 Epoch 9/10 155/155 [==============================] - 38s 242ms/step - loss: 0.1428 - accuracy: 0.9586 - val_loss: 0.1606 - val_accuracy: 0.9560 Epoch 10/10 155/155 [==============================] - 35s 229ms/step - loss: 0.1195 - accuracy: 0.9663 - val_loss: 0.1309 - val_accuracy: 0.9720
In [ ]:
Copied!
model_1.evaluate(test_data)
model_1.evaluate(test_data)
8/8 [==============================] - 1s 128ms/step - loss: 0.1758 - accuracy: 0.9680
Out[ ]:
[0.17583806812763214, 0.9679999947547913]
Compared historys with the original model¶
In [ ]:
Copied!
compare_historys(history_1, history_1_fine_tune_1)
compare_historys(history_1, history_1_fine_tune_1)
In [ ]:
Copied!
model_1.save("/content/drive/MyDrive/Butterfly-Recognition/Model_2")
model_1.save("/content/drive/MyDrive/Butterfly-Recognition/Model_2")
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument. category=CustomMaskWarning)
INFO:tensorflow:Assets written to: /content/drive/MyDrive/Butterfly-Recognition/Model_2/assets
Insights into the best model¶
In [ ]:
Copied!
label_val = []
for elem in test_data.filepaths:
for cl in range(50):
if elem.split("/")[4] == class_names[cl]:
label_val.append(cl)
label_val = []
for elem in test_data.filepaths:
for cl in range(50):
if elem.split("/")[4] == class_names[cl]:
label_val.append(cl)
In [ ]:
Copied!
label_val[5:10]
label_val[5:10]
Out[ ]:
[1, 1, 1, 1, 1]
In [ ]:
Copied!
# Getting prediction probabilites
pred_prob = model_1.predict(test_data)
pred_prob
# Getting prediction probabilites
pred_prob = model_1.predict(test_data)
pred_prob
Out[ ]:
array([[8.1247598e-01, 1.9608064e-05, 8.4864783e-05, ..., 3.2571238e-04,
1.7534398e-05, 9.8661276e-06],
[5.8847833e-01, 2.0846370e-05, 2.4964868e-06, ..., 2.0753338e-05,
2.3687000e-05, 6.9808941e-08],
[9.7818691e-01, 1.7820315e-06, 3.9221965e-07, ..., 5.4445458e-08,
1.9965878e-04, 9.5505243e-07],
...,
[2.0196762e-06, 1.0442600e-04, 1.8454040e-05, ..., 1.2071504e-07,
8.4692489e-05, 9.8753589e-01],
[2.8213247e-09, 1.4057372e-07, 3.0933300e-08, ..., 2.7334593e-08,
3.7891237e-04, 9.9538982e-01],
[1.5570292e-08, 4.8477432e-06, 2.1701456e-07, ..., 1.0687562e-09,
2.0701880e-06, 9.9944359e-01]], dtype=float32)
In [ ]:
Copied!
y_pred = pred_prob.argmax(axis=1)
y_pred[:50]
y_pred = pred_prob.argmax(axis=1)
y_pred[:50]
Out[ ]:
array([ 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 3, 3,
3, 3, 3, 4, 19, 4, 4, 4, 5, 5, 5, 5, 5, 6, 6, 6, 6,
6, 7, 7, 7, 31, 7, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9])
In [ ]:
Copied!
y_true = np.array(label_val)
y_true[:50]
y_true = np.array(label_val)
y_true[:50]
Out[ ]:
array([0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 4, 4,
4, 4, 4, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 8, 8, 8, 8,
8, 9, 9, 9, 9, 9])
Confusion Matrix¶
In [ ]:
Copied!
make_confusion_matrix(y_true=y_true,
y_pred=y_pred,
classes=class_names,
figsize=(50, 50),
text_size=20,
norm=False)
make_confusion_matrix(y_true=y_true,
y_pred=y_pred,
classes=class_names,
figsize=(50, 50),
text_size=20,
norm=False)
Classification Report¶
In [ ]:
Copied!
print(classification_report(y_true, y_pred))
print(classification_report(y_true, y_pred))
precision recall f1-score support
0 1.00 1.00 1.00 5
1 1.00 1.00 1.00 5
2 1.00 1.00 1.00 5
3 1.00 1.00 1.00 5
4 1.00 0.80 0.89 5
5 1.00 1.00 1.00 5
6 1.00 1.00 1.00 5
7 1.00 0.80 0.89 5
8 1.00 1.00 1.00 5
9 1.00 1.00 1.00 5
10 1.00 1.00 1.00 5
11 1.00 0.80 0.89 5
12 1.00 1.00 1.00 5
13 1.00 0.80 0.89 5
14 1.00 1.00 1.00 5
15 1.00 1.00 1.00 5
16 0.83 1.00 0.91 5
17 1.00 1.00 1.00 5
18 1.00 1.00 1.00 5
19 0.83 1.00 0.91 5
20 1.00 0.80 0.89 5
21 1.00 1.00 1.00 5
22 1.00 1.00 1.00 5
23 1.00 1.00 1.00 5
24 1.00 1.00 1.00 5
25 1.00 1.00 1.00 5
26 1.00 1.00 1.00 5
27 1.00 1.00 1.00 5
28 0.83 1.00 0.91 5
29 1.00 1.00 1.00 5
30 1.00 1.00 1.00 5
31 0.83 1.00 0.91 5
32 0.83 1.00 0.91 5
33 0.75 0.60 0.67 5
34 0.83 1.00 0.91 5
35 1.00 1.00 1.00 5
36 1.00 1.00 1.00 5
37 1.00 1.00 1.00 5
38 1.00 1.00 1.00 5
39 1.00 1.00 1.00 5
40 1.00 1.00 1.00 5
41 1.00 0.80 0.89 5
42 1.00 1.00 1.00 5
43 1.00 1.00 1.00 5
44 1.00 1.00 1.00 5
45 0.83 1.00 0.91 5
46 1.00 1.00 1.00 5
47 1.00 1.00 1.00 5
48 1.00 1.00 1.00 5
49 1.00 1.00 1.00 5
accuracy 0.97 250
macro avg 0.97 0.97 0.97 250
weighted avg 0.97 0.97 0.97 250
In [ ]:
Copied!
# Get a dictionary of the classification report
classification_report_dict = classification_report(y_true, y_pred, output_dict=True)
classification_report_dict
# Get a dictionary of the classification report
classification_report_dict = classification_report(y_true, y_pred, output_dict=True)
classification_report_dict
Out[ ]:
{'0': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'1': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'10': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'11': {'f1-score': 0.888888888888889,
'precision': 1.0,
'recall': 0.8,
'support': 5},
'12': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'13': {'f1-score': 0.888888888888889,
'precision': 1.0,
'recall': 0.8,
'support': 5},
'14': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'15': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'16': {'f1-score': 0.9090909090909091,
'precision': 0.8333333333333334,
'recall': 1.0,
'support': 5},
'17': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'18': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'19': {'f1-score': 0.9090909090909091,
'precision': 0.8333333333333334,
'recall': 1.0,
'support': 5},
'2': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'20': {'f1-score': 0.888888888888889,
'precision': 1.0,
'recall': 0.8,
'support': 5},
'21': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'22': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'23': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'24': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'25': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'26': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'27': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'28': {'f1-score': 0.9090909090909091,
'precision': 0.8333333333333334,
'recall': 1.0,
'support': 5},
'29': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'3': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'30': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'31': {'f1-score': 0.9090909090909091,
'precision': 0.8333333333333334,
'recall': 1.0,
'support': 5},
'32': {'f1-score': 0.9090909090909091,
'precision': 0.8333333333333334,
'recall': 1.0,
'support': 5},
'33': {'f1-score': 0.6666666666666665,
'precision': 0.75,
'recall': 0.6,
'support': 5},
'34': {'f1-score': 0.9090909090909091,
'precision': 0.8333333333333334,
'recall': 1.0,
'support': 5},
'35': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'36': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'37': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'38': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'39': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'4': {'f1-score': 0.888888888888889,
'precision': 1.0,
'recall': 0.8,
'support': 5},
'40': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'41': {'f1-score': 0.888888888888889,
'precision': 1.0,
'recall': 0.8,
'support': 5},
'42': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'43': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'44': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'45': {'f1-score': 0.9090909090909091,
'precision': 0.8333333333333334,
'recall': 1.0,
'support': 5},
'46': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'47': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'48': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'49': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'5': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'6': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'7': {'f1-score': 0.888888888888889,
'precision': 1.0,
'recall': 0.8,
'support': 5},
'8': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'9': {'f1-score': 1.0, 'precision': 1.0, 'recall': 1.0, 'support': 5},
'accuracy': 0.968,
'macro avg': {'f1-score': 0.9672727272727272,
'precision': 0.9716666666666667,
'recall': 0.968,
'support': 250},
'weighted avg': {'f1-score': 0.9672727272727273,
'precision': 0.9716666666666668,
'recall': 0.968,
'support': 250}}
In [ ]:
Copied!
# Create empty dictionary
class_f1_scores = {}
# Loop through classification report items
for k, v in classification_report_dict.items():
if k == "accuracy": # stop once we get to accuracy key
break
else:
# Append class names and f1-scores to new dictionary
class_f1_scores[class_names[int(k)]] = v["f1-score"]
class_f1_scores
# Create empty dictionary
class_f1_scores = {}
# Loop through classification report items
for k, v in classification_report_dict.items():
if k == "accuracy": # stop once we get to accuracy key
break
else:
# Append class names and f1-scores to new dictionary
class_f1_scores[class_names[int(k)]] = v["f1-score"]
class_f1_scores
Out[ ]:
{'adonis': 1.0,
'american snoot': 1.0,
'an 88': 1.0,
'banded peacock': 1.0,
'beckers white': 0.888888888888889,
'black hairstreak': 1.0,
'cabbage white': 1.0,
'chestnut': 0.888888888888889,
'clodius parnassian': 1.0,
'clouded sulphur': 1.0,
'copper tail': 1.0,
'crecent': 0.888888888888889,
'crimson patch': 1.0,
'eastern coma': 0.888888888888889,
'gold banded': 1.0,
'great eggfly': 1.0,
'grey hairstreak': 0.9090909090909091,
'indra swallow': 1.0,
'julia': 1.0,
'large marble': 0.9090909090909091,
'malachite': 0.888888888888889,
'mangrove skipper': 1.0,
'metalmark': 1.0,
'monarch': 1.0,
'morning cloak': 1.0,
'orange oakleaf': 1.0,
'orange tip': 1.0,
'orchard swallow': 1.0,
'painted lady': 0.9090909090909091,
'paper kite': 1.0,
'peacock': 1.0,
'pine white': 0.9090909090909091,
'pipevine swallow': 0.9090909090909091,
'purple hairstreak': 0.6666666666666665,
'question mark': 0.9090909090909091,
'red admiral': 1.0,
'red spotted purple': 1.0,
'scarce swallow': 1.0,
'silver spot skipper': 1.0,
'sixspot burnet': 1.0,
'skipper': 1.0,
'sootywing': 0.888888888888889,
'southern dogface': 1.0,
'straited queen': 1.0,
'two barred flasher': 1.0,
'ulyses': 0.9090909090909091,
'viceroy': 1.0,
'wood satyr': 1.0,
'yellow swallow tail': 1.0,
'zebra long wing': 1.0}
In [ ]:
Copied!
# Turn f1-scores into dataframe for visualization
f1_scores = pd.DataFrame({"class_name": list(class_f1_scores.keys()),
"f1-score": list(class_f1_scores.values())}).sort_values("f1-score", ascending=False)
f1_scores
# Turn f1-scores into dataframe for visualization
f1_scores = pd.DataFrame({"class_name": list(class_f1_scores.keys()),
"f1-score": list(class_f1_scores.values())}).sort_values("f1-score", ascending=False)
f1_scores
Out[ ]:
| class_name | f1-score | |
|---|---|---|
| 0 | adonis | 1.000000 |
| 37 | scarce swallow | 1.000000 |
| 24 | morning cloak | 1.000000 |
| 1 | american snoot | 1.000000 |
| 26 | orange tip | 1.000000 |
| 27 | orchard swallow | 1.000000 |
| 29 | paper kite | 1.000000 |
| 30 | peacock | 1.000000 |
| 35 | red admiral | 1.000000 |
| 36 | red spotted purple | 1.000000 |
| 38 | silver spot skipper | 1.000000 |
| 22 | metalmark | 1.000000 |
| 39 | sixspot burnet | 1.000000 |
| 40 | skipper | 1.000000 |
| 42 | southern dogface | 1.000000 |
| 43 | straited queen | 1.000000 |
| 44 | two barred flasher | 1.000000 |
| 46 | viceroy | 1.000000 |
| 47 | wood satyr | 1.000000 |
| 48 | yellow swallow tail | 1.000000 |
| 23 | monarch | 1.000000 |
| 25 | orange oakleaf | 1.000000 |
| 21 | mangrove skipper | 1.000000 |
| 10 | copper tail | 1.000000 |
| 2 | an 88 | 1.000000 |
| 3 | banded peacock | 1.000000 |
| 5 | black hairstreak | 1.000000 |
| 6 | cabbage white | 1.000000 |
| 8 | clodius parnassian | 1.000000 |
| 9 | clouded sulphur | 1.000000 |
| 12 | crimson patch | 1.000000 |
| 49 | zebra long wing | 1.000000 |
| 18 | julia | 1.000000 |
| 17 | indra swallow | 1.000000 |
| 14 | gold banded | 1.000000 |
| 15 | great eggfly | 1.000000 |
| 32 | pipevine swallow | 0.909091 |
| 19 | large marble | 0.909091 |
| 45 | ulyses | 0.909091 |
| 16 | grey hairstreak | 0.909091 |
| 31 | pine white | 0.909091 |
| 34 | question mark | 0.909091 |
| 28 | painted lady | 0.909091 |
| 41 | sootywing | 0.888889 |
| 7 | chestnut | 0.888889 |
| 11 | crecent | 0.888889 |
| 20 | malachite | 0.888889 |
| 13 | eastern coma | 0.888889 |
| 4 | beckers white | 0.888889 |
| 33 | purple hairstreak | 0.666667 |
Plotting F1 scores¶
In [ ]:
Copied!
# plotting F1 scores
fig, ax = plt.subplots(figsize=(10, 10))
scores = ax.barh(range(len(f1_scores)), f1_scores["f1-score"].values)
ax.set_yticks(range(len(f1_scores)))
ax.set_yticklabels(list(f1_scores["class_name"]))
ax.set_xlabel("f1-score")
ax.set_title("F1-Scores for 5 Different Classes")
ax.invert_yaxis(); # reverse the order
def autolabel(rects):
"""
Attach a text label above each bar displaying its height (it's value).
"""
for rect in rects:
width = rect.get_width()
ax.text(1.03*width, rect.get_y() + rect.get_height()/1.5,
f"{width:.2f}",
ha='center', va='bottom')
autolabel(scores)
# plotting F1 scores
fig, ax = plt.subplots(figsize=(10, 10))
scores = ax.barh(range(len(f1_scores)), f1_scores["f1-score"].values)
ax.set_yticks(range(len(f1_scores)))
ax.set_yticklabels(list(f1_scores["class_name"]))
ax.set_xlabel("f1-score")
ax.set_title("F1-Scores for 5 Different Classes")
ax.invert_yaxis(); # reverse the order
def autolabel(rects):
"""
Attach a text label above each bar displaying its height (it's value).
"""
for rect in rects:
width = rect.get_width()
ax.text(1.03*width, rect.get_y() + rect.get_height()/1.5,
f"{width:.2f}",
ha='center', va='bottom')
autolabel(scores)
Predictions on the model¶
In [ ]:
Copied!
plt.figure(figsize=(17, 12))
for i in range(9):
# Choose a random image from a random class
filepath = random.choice(test_paths)
class_name = filepath.split("/")[4]
# Load the image and make predictions
img = load_and_preprocess(filepath)
pred_prob = model_1.predict(tf.expand_dims(img, axis=0)) # model accepts tensors of shape [None, 224, 224, 3]
pred_class = class_names[pred_prob.argmax()] # find the predicted class
# Plot the image(s)
plt.subplot(3, 3, i+1)
plt.imshow(img/255.)
if class_name == pred_class: # Change the color of text based on whether prediction is right or wrong
title_color = "g"
else:
title_color = "r"
plt.title(f"actual: {class_name}, pred: {pred_class}, \nprob: {pred_prob.max():.2f}%", c=title_color)
plt.axis(False);
plt.figure(figsize=(17, 12))
for i in range(9):
# Choose a random image from a random class
filepath = random.choice(test_paths)
class_name = filepath.split("/")[4]
# Load the image and make predictions
img = load_and_preprocess(filepath)
pred_prob = model_1.predict(tf.expand_dims(img, axis=0)) # model accepts tensors of shape [None, 224, 224, 3]
pred_class = class_names[pred_prob.argmax()] # find the predicted class
# Plot the image(s)
plt.subplot(3, 3, i+1)
plt.imshow(img/255.)
if class_name == pred_class: # Change the color of text based on whether prediction is right or wrong
title_color = "g"
else:
title_color = "r"
plt.title(f"actual: {class_name}, pred: {pred_class}, \nprob: {pred_prob.max():.2f}%", c=title_color)
plt.axis(False);
In [ ]:
Copied!
df_pred = pd.DataFrame({"filename" : test_paths,
"label": test_labels,
"label_val": label_val,
"pred": y_pred})
df_pred = pd.DataFrame({"filename" : test_paths,
"label": test_labels,
"label_val": label_val,
"pred": y_pred})
In [ ]:
Copied!
df_pred.head()
df_pred.head()
Out[ ]:
| filename | label | label_val | pred | |
|---|---|---|---|---|
| 0 | /content/butterflies/test/adonis/1.jpg | adonis | 0 | 0 |
| 1 | /content/butterflies/test/adonis/2.jpg | adonis | 0 | 0 |
| 2 | /content/butterflies/test/adonis/3.jpg | adonis | 0 | 0 |
| 3 | /content/butterflies/test/adonis/4.jpg | adonis | 0 | 0 |
| 4 | /content/butterflies/test/adonis/5.jpg | adonis | 0 | 0 |
In [ ]:
Copied!
pred_probs = model_1.predict(test_data)
pred_probs
pred_probs = model_1.predict(test_data)
pred_probs
Out[ ]:
array([[8.1247598e-01, 1.9608064e-05, 8.4864783e-05, ..., 3.2571238e-04,
1.7534398e-05, 9.8661276e-06],
[5.8847833e-01, 2.0846370e-05, 2.4964868e-06, ..., 2.0753338e-05,
2.3687000e-05, 6.9808941e-08],
[9.7818691e-01, 1.7820315e-06, 3.9221965e-07, ..., 5.4445458e-08,
1.9965878e-04, 9.5505243e-07],
...,
[2.0196762e-06, 1.0442600e-04, 1.8454040e-05, ..., 1.2071504e-07,
8.4692489e-05, 9.8753589e-01],
[2.8213247e-09, 1.4057372e-07, 3.0933300e-08, ..., 2.7334593e-08,
3.7891237e-04, 9.9538982e-01],
[1.5570292e-08, 4.8477432e-06, 2.1701456e-07, ..., 1.0687562e-09,
2.0701880e-06, 9.9944359e-01]], dtype=float32)
In [ ]:
Copied!
pred_df = pd.DataFrame({"img_path": test_paths,
"y_true": y_true,
"y_pred": y_pred,
"pred_conf": pred_probs.max(axis=1), # get the maximum prediction probability value
"y_true_classname": [class_names[i] for i in y_true],
"y_pred_classname": [class_names[i] for i in y_pred]})
pred_df.head()
pred_df = pd.DataFrame({"img_path": test_paths,
"y_true": y_true,
"y_pred": y_pred,
"pred_conf": pred_probs.max(axis=1), # get the maximum prediction probability value
"y_true_classname": [class_names[i] for i in y_true],
"y_pred_classname": [class_names[i] for i in y_pred]})
pred_df.head()
Out[ ]:
| img_path | y_true | y_pred | pred_conf | y_true_classname | y_pred_classname | |
|---|---|---|---|---|---|---|
| 0 | /content/butterflies/test/adonis/1.jpg | 0 | 0 | 0.812476 | adonis | adonis |
| 1 | /content/butterflies/test/adonis/2.jpg | 0 | 0 | 0.588478 | adonis | adonis |
| 2 | /content/butterflies/test/adonis/3.jpg | 0 | 0 | 0.978187 | adonis | adonis |
| 3 | /content/butterflies/test/adonis/4.jpg | 0 | 0 | 0.803290 | adonis | adonis |
| 4 | /content/butterflies/test/adonis/5.jpg | 0 | 0 | 0.460714 | adonis | adonis |
In [ ]:
Copied!
# check for pred
pred_df["pred_correct"] = pred_df["y_true"] == pred_df["y_pred"]
pred_df.head()
# check for pred
pred_df["pred_correct"] = pred_df["y_true"] == pred_df["y_pred"]
pred_df.head()
Out[ ]:
| img_path | y_true | y_pred | pred_conf | y_true_classname | y_pred_classname | pred_correct | |
|---|---|---|---|---|---|---|---|
| 0 | /content/butterflies/test/adonis/1.jpg | 0 | 0 | 0.812476 | adonis | adonis | True |
| 1 | /content/butterflies/test/adonis/2.jpg | 0 | 0 | 0.588478 | adonis | adonis | True |
| 2 | /content/butterflies/test/adonis/3.jpg | 0 | 0 | 0.978187 | adonis | adonis | True |
| 3 | /content/butterflies/test/adonis/4.jpg | 0 | 0 | 0.803290 | adonis | adonis | True |
| 4 | /content/butterflies/test/adonis/5.jpg | 0 | 0 | 0.460714 | adonis | adonis | True |
In [ ]:
Copied!
# Get the top wrong examples
top_wrong = pred_df[pred_df["pred_correct"] == False].sort_values("pred_conf", ascending=False)[:10]
top_wrong.head(20)
# Get the top wrong examples
top_wrong = pred_df[pred_df["pred_correct"] == False].sort_values("pred_conf", ascending=False)[:10]
top_wrong.head(20)
Out[ ]:
| img_path | y_true | y_pred | pred_conf | y_true_classname | y_pred_classname | pred_correct | |
|---|---|---|---|---|---|---|---|
| 55 | /content/butterflies/test/crecent/1.jpg | 11 | 28 | 0.829781 | crecent | painted lady | False |
| 169 | /content/butterflies/test/purple hairstreak/5.jpg | 33 | 45 | 0.725551 | purple hairstreak | ulyses | False |
| 21 | /content/butterflies/test/beckers white/2.jpg | 4 | 19 | 0.716408 | beckers white | large marble | False |
| 100 | /content/butterflies/test/malachite/1.jpg | 20 | 32 | 0.704015 | malachite | pipevine swallow | False |
| 168 | /content/butterflies/test/purple hairstreak/4.jpg | 33 | 16 | 0.641446 | purple hairstreak | grey hairstreak | False |
| 207 | /content/butterflies/test/sootywing/3.jpg | 41 | 33 | 0.622935 | sootywing | purple hairstreak | False |
| 69 | /content/butterflies/test/eastern coma/5.jpg | 13 | 34 | 0.460990 | eastern coma | question mark | False |
| 38 | /content/butterflies/test/chestnut/4.jpg | 7 | 31 | 0.410461 | chestnut | pine white | False |
In [ ]:
Copied!
filepaths = np.array(top_wrong["img_path"])
true = np.array(top_wrong["y_true_classname"])
pred = np.array(top_wrong["y_pred_classname"])
prob = np.array(top_wrong["pred_conf"])
filepaths = np.array(top_wrong["img_path"])
true = np.array(top_wrong["y_true_classname"])
pred = np.array(top_wrong["y_pred_classname"])
prob = np.array(top_wrong["pred_conf"])
Wrongly predicted classes¶
In [ ]:
Copied!
# Plotting wrong predictions with insights
plt.figure(figsize=(17, 12))
for i in range(len(filepaths)):
img = load_and_preprocess(filepaths[i])
plt.subplot(3, 3, i+1)
plt.axis(False)
plt.imshow(img / 255.)
plt.title(f"Actual: {true[i]}, Pred: {pred[i]}, \nprob: {prob[i]:.2f}%", color="red");
# Plotting wrong predictions with insights
plt.figure(figsize=(17, 12))
for i in range(len(filepaths)):
img = load_and_preprocess(filepaths[i])
plt.subplot(3, 3, i+1)
plt.axis(False)
plt.imshow(img / 255.)
plt.title(f"Actual: {true[i]}, Pred: {pred[i]}, \nprob: {prob[i]:.2f}%", color="red");
In [ ]:
Copied!