Fruit-Classification
Data mount and Path Set-up for the Project on the Google Drive
from google.colab import drive
drive.mount('/content/gdrive')
Mounted at /content/gdrive
# Data Path Setup
data_path = "/content/gdrive/MyDrive/'Colab Notebooks'/hci575_data/"
data_path
"/content/gdrive/MyDrive/'Colab Notebooks'/hci575_data/"
!pip install -q kaggle
from google.colab import files
files.upload()
<input type="file" id="files-604af184-517e-4ee2-b275-170996836ab0" name="files[]" multiple disabled
style="border:none" />
<output id="result-604af184-517e-4ee2-b275-170996836ab0">
Upload widget is only available when the cell has been executed in the
current browser session. Please rerun this cell to enable.
</output>
<script>// Copyright 2017 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. // You may obtain a copy of the License at // // http://www.apache.org/licenses/LICENSE-2.0 // // Unless required by applicable law or agreed to in writing, software // distributed under the License is distributed on an "AS IS" BASIS, // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. // See the License for the specific language governing permissions and // limitations under the License.
/**
- @fileoverview Helpers for google.colab Python module. */ (function(scope) { function span(text, styleAttributes = {}) { const element = document.createElement(‘span’); element.textContent = text; for (const key of Object.keys(styleAttributes)) { element.style[key] = styleAttributes[key]; } return element; }
// Max number of bytes which will be uploaded at a time. const MAX_PAYLOAD_SIZE = 100 * 1024;
function _uploadFiles(inputId, outputId) { const steps = uploadFilesStep(inputId, outputId); const outputElement = document.getElementById(outputId); // Cache steps on the outputElement to make it available for the next call // to uploadFilesContinue from Python. outputElement.steps = steps;
return _uploadFilesContinue(outputId); }
// This is roughly an async generator (not supported in the browser yet), // where there are multiple asynchronous steps and the Python side is going // to poll for completion of each step. // This uses a Promise to block the python side on completion of each step, // then passes the result of the previous step as the input to the next step. function _uploadFilesContinue(outputId) { const outputElement = document.getElementById(outputId); const steps = outputElement.steps;
const next = steps.next(outputElement.lastPromiseValue); return Promise.resolve(next.value.promise).then((value) => { // Cache the last promise value to make it available to the next // step of the generator. outputElement.lastPromiseValue = value; return next.value.response; }); }
/**
- Generator function which is called between each async step of the upload
- process.
- @param {string} inputId Element ID of the input file picker element.
- @param {string} outputId Element ID of the output display.
- @return {!Iterable<!Object>} Iterable of next steps. / function uploadFilesStep(inputId, outputId) { const inputElement = document.getElementById(inputId); inputElement.disabled = false;
const outputElement = document.getElementById(outputId); outputElement.innerHTML = ‘’;
const pickedPromise = new Promise((resolve) => { inputElement.addEventListener(‘change’, (e) => { resolve(e.target.files); }); });
const cancel = document.createElement(‘button’); inputElement.parentElement.appendChild(cancel); cancel.textContent = ‘Cancel upload’; const cancelPromise = new Promise((resolve) => { cancel.onclick = () => { resolve(null); }; });
// Wait for the user to pick the files. const files = yield { promise: Promise.race([pickedPromise, cancelPromise]), response: { action: ‘starting’, } };
cancel.remove();
// Disable the input element since further picks are not allowed. inputElement.disabled = true;
if (!files) { return { response: { action: ‘complete’, } }; }
for (const file of files) {
const li = document.createElement(‘li’);
li.append(span(file.name, {fontWeight: ‘bold’}));
li.append(span(
(${file.type || 'n/a'}) - ${file.size} bytes, +
last modified: ${
file.lastModifiedDate ? file.lastModifiedDate.toLocaleDateString() :
'n/a'} - ));
const percent = span(‘0% done’);
li.appendChild(percent);
outputElement.appendChild(li);
const fileDataPromise = new Promise((resolve) => {
const reader = new FileReader();
reader.onload = (e) => {
resolve(e.target.result);
};
reader.readAsArrayBuffer(file);
});
// Wait for the data to be ready.
let fileData = yield {
promise: fileDataPromise,
response: {
action: 'continue',
}
};
// Use a chunked sending to avoid message size limits. See b/62115660.
let position = 0;
do {
const length = Math.min(fileData.byteLength - position, MAX_PAYLOAD_SIZE);
const chunk = new Uint8Array(fileData, position, length);
position += length;
const base64 = btoa(String.fromCharCode.apply(null, chunk));
yield {
response: {
action: 'append',
file: file.name,
data: base64,
},
};
let percentDone = fileData.byteLength === 0 ?
100 :
Math.round((position / fileData.byteLength) * 100);
percent.textContent = `${percentDone}% done`;
} while (position < fileData.byteLength); }
// All done. yield { response: { action: ‘complete’, } }; }
scope.google = scope.google || {}; scope.google.colab = scope.google.colab || {}; scope.google.colab._files = { _uploadFiles, _uploadFilesContinue, }; })(self); </script>
Saving kaggle.json to kaggle.json
{'kaggle.json': b'{"username":"kkumarr","key":"22035d1ea9e36d000bf46a1abca21e46"}'}
!mkdir ~/.kaggle
!cp kaggle.json ~/.kaggle/
!chmod 600 ~/.kaggle/kaggle.json
!kaggle datasets list
ref title size lastUpdated downloadCount voteCount usabilityRating
---------------------------------------------------------- ------------------------------------------- ----- ------------------- ------------- --------- ---------------
arnabchaki/data-science-salaries-2023 Data Science Salaries 2023 💸 25KB 2023-04-13 09:55:16 5820 141 1.0
salvatorerastelli/spotify-and-youtube Spotify and Youtube 9MB 2023-03-20 15:43:25 11456 400 1.0
iammustafatz/diabetes-prediction-dataset Diabetes prediction dataset 734KB 2023-04-08 06:11:45 2986 51 1.0
shawkyelgendy/furniture-price-prediction Furniture Price Prediction 81KB 2023-04-17 09:35:17 957 27 1.0
arnabchaki/indian-restaurants-2023 Indian Restaurants 2023 🍲 139KB 2023-04-27 09:57:57 580 23 1.0
kapturovalexander/ferrari-and-tesla-share-prices-2015-2023 Ferrari and Tesla. Share prices (2015-2023) 82KB 2023-04-30 11:28:39 594 29 1.0
desalegngeb/students-exam-scores Students Exam Scores: Extended Dataset 695KB 2023-04-14 00:15:38 2157 57 1.0
erdemtaha/cancer-data Cancer Data 49KB 2023-03-22 07:57:00 5893 116 1.0
lokeshparab/amazon-products-dataset Amazon Products Sales Dataset 2023 80MB 2023-03-26 10:45:19 6107 131 1.0
cuecacuela/england-football-matches England Football Matches 389KB 2023-04-25 16:08:27 469 25 1.0
matarrgaye/uk-consumer-trends-current-price UK Consumer Trends: 1997 - 2022, Quarterly 293KB 2023-04-26 14:49:10 421 21 0.88235295
harshsingh2209/supply-chain-analysis Supply Chain Analysis 9KB 2023-04-22 23:32:27 436 23 1.0
ulrikthygepedersen/fastfood-nutrition Fastfood Nutrition 12KB 2023-03-21 10:02:41 4799 91 1.0
priyanshusethi/minecraft-piracy-dataset Minecraft Piracy Dataset 30KB 2023-04-20 14:14:10 330 32 1.0
ppb00x/credit-risk-customers credit_risk_customers 18KB 2023-04-12 08:28:28 2007 49 1.0
evangower/premier-league-2022-2023 Premier League 2022-2023 7KB 2023-04-14 19:45:22 1553 38 1.0
arnabchaki/popular-video-games-1980-2023 Popular Video Games 1980 - 2023 🎮 1MB 2023-03-23 16:16:51 4981 133 1.0
dansbecker/melbourne-housing-snapshot Melbourne Housing Snapshot 451KB 2018-06-05 12:52:24 114072 1252 0.7058824
shubhammeshram579/house Housing price prediction 12KB 2023-04-19 09:37:25 1094 35 1.0
r1chardson/the-world-university-rankings-2011-2023 THE World University Rankings 2011-2023 1MB 2023-04-03 12:43:37 2998 65 1.0
! kaggle datasets download -d moltean/fruits
Downloading fruits.zip to /content/gdrive/MyDrive/Colab Notebooks/hci575_data
99% 1.27G/1.28G [00:12<00:00, 104MB/s]
100% 1.28G/1.28G [00:12<00:00, 114MB/s]
!unzip -q fruits.zip
Data preprocessing for Training Deep Learning Model
!pip install tqdm
Collecting tqdm
Downloading tqdm-4.65.0-py3-none-any.whl (77 kB)
[2K [90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m77.1/77.1 kB[0m [31m1.2 MB/s[0m eta [36m0:00:00[0ma [36m0:00:01[0m
[?25hInstalling collected packages: tqdm
Successfully installed tqdm-4.65.0
!pip install numpy
Collecting numpy
Downloading numpy-1.24.3-cp38-cp38-macosx_10_9_x86_64.whl (19.8 MB)
[2K [90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m19.8/19.8 MB[0m [31m6.0 MB/s[0m eta [36m0:00:00[0m00:01[0m00:01[0mm
[?25hInstalling collected packages: numpy
Successfully installed numpy-1.24.3
!pip install matplotlib
Collecting matplotlib
Downloading matplotlib-3.7.1-cp38-cp38-macosx_10_12_x86_64.whl (7.4 MB)
[2K [90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m7.4/7.4 MB[0m [31m9.5 MB/s[0m eta [36m0:00:00[0m00:01[0m00:01[0m
[?25hCollecting pyparsing>=2.3.1
Using cached pyparsing-3.0.9-py3-none-any.whl (98 kB)
Collecting cycler>=0.10
Using cached cycler-0.11.0-py3-none-any.whl (6.4 kB)
Requirement already satisfied: importlib-resources>=3.2.0 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib) (5.2.0)
Collecting pillow>=6.2.0
Downloading Pillow-9.5.0-cp38-cp38-macosx_10_10_x86_64.whl (3.4 MB)
[2K [90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m3.4/3.4 MB[0m [31m11.2 MB/s[0m eta [36m0:00:00[0ma [36m0:00:01[0m
[?25hRequirement already satisfied: numpy>=1.20 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib) (1.24.3)
Collecting kiwisolver>=1.0.1
Using cached kiwisolver-1.4.4-cp38-cp38-macosx_10_9_x86_64.whl (65 kB)
Collecting contourpy>=1.0.1
Downloading contourpy-1.0.7-cp38-cp38-macosx_10_9_x86_64.whl (243 kB)
[2K [90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m244.0/244.0 kB[0m [31m8.3 MB/s[0m eta [36m0:00:00[0m
[?25hRequirement already satisfied: packaging>=20.0 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib) (23.0)
Collecting fonttools>=4.22.0
Downloading fonttools-4.39.3-py3-none-any.whl (1.0 MB)
[2K [90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m1.0/1.0 MB[0m [31m10.8 MB/s[0m eta [36m0:00:00[0ma [36m0:00:01[0m
[?25hRequirement already satisfied: python-dateutil>=2.7 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib) (2.8.2)
Requirement already satisfied: zipp>=3.1.0 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from importlib-resources>=3.2.0->matplotlib) (3.11.0)
Requirement already satisfied: six>=1.5 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from python-dateutil>=2.7->matplotlib) (1.16.0)
Installing collected packages: pyparsing, pillow, kiwisolver, fonttools, cycler, contourpy, matplotlib
Successfully installed contourpy-1.0.7 cycler-0.11.0 fonttools-4.39.3 kiwisolver-1.4.4 matplotlib-3.7.1 pillow-9.5.0 pyparsing-3.0.9
# Packages
# import pandas as pd
import numpy as np
# import cv2
import PIL
import os
import matplotlib.pyplot as plt
from tqdm import tqdm_notebook
import random
import shutil
!pip install seaborn
Collecting seaborn
Downloading seaborn-0.12.2-py3-none-any.whl (293 kB)
[2K [90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m293.3/293.3 kB[0m [31m4.2 MB/s[0m eta [36m0:00:00[0ma [36m0:00:01[0m
[?25hRequirement already satisfied: matplotlib!=3.6.1,>=3.1 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from seaborn) (3.7.1)
Collecting pandas>=0.25
Downloading pandas-2.0.1-cp38-cp38-macosx_10_9_x86_64.whl (11.6 MB)
[2K [90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m11.6/11.6 MB[0m [31m36.8 MB/s[0m eta [36m0:00:00[0m00:01[0m00:01[0m
[?25hRequirement already satisfied: numpy!=1.24.0,>=1.17 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from seaborn) (1.24.3)
Requirement already satisfied: pillow>=6.2.0 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib!=3.6.1,>=3.1->seaborn) (9.5.0)
Requirement already satisfied: contourpy>=1.0.1 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib!=3.6.1,>=3.1->seaborn) (1.0.7)
Requirement already satisfied: packaging>=20.0 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib!=3.6.1,>=3.1->seaborn) (23.0)
Requirement already satisfied: importlib-resources>=3.2.0 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib!=3.6.1,>=3.1->seaborn) (5.2.0)
Requirement already satisfied: python-dateutil>=2.7 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib!=3.6.1,>=3.1->seaborn) (2.8.2)
Requirement already satisfied: fonttools>=4.22.0 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib!=3.6.1,>=3.1->seaborn) (4.39.3)
Requirement already satisfied: pyparsing>=2.3.1 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib!=3.6.1,>=3.1->seaborn) (3.0.9)
Requirement already satisfied: cycler>=0.10 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib!=3.6.1,>=3.1->seaborn) (0.11.0)
Requirement already satisfied: kiwisolver>=1.0.1 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from matplotlib!=3.6.1,>=3.1->seaborn) (1.4.4)
Collecting tzdata>=2022.1
Downloading tzdata-2023.3-py2.py3-none-any.whl (341 kB)
[2K [90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m341.8/341.8 kB[0m [31m34.4 MB/s[0m eta [36m0:00:00[0m
[?25hRequirement already satisfied: pytz>=2020.1 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from pandas>=0.25->seaborn) (2022.7)
Requirement already satisfied: zipp>=3.1.0 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from importlib-resources>=3.2.0->matplotlib!=3.6.1,>=3.1->seaborn) (3.11.0)
Requirement already satisfied: six>=1.5 in /Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages (from python-dateutil>=2.7->matplotlib!=3.6.1,>=3.1->seaborn) (1.16.0)
Installing collected packages: tzdata, pandas, seaborn
Successfully installed pandas-2.0.1 seaborn-0.12.2 tzdata-2023.3
# Path Set up for Fruits class label directories
moldy_fruits = "/content/gdrive/MyDrive/Colab Notebooks/hci575_data/moldy_fruits"
overripe_fruits = "/content/gdrive/MyDrive/Colab Notebooks/hci575_data/overripe_fruits"
ripe_fruits = "/content/gdrive/MyDrive/Colab Notebooks/hci575_data/ripe_fruits"
unripe_fruits = "/content/gdrive/MyDrive/Colab Notebooks/hci575_data/unripe_fruits"
cd /content/gdrive/MyDrive/'Colab Notebooks'/hci575_data/
/content/gdrive/MyDrive/Colab Notebooks/hci575_data
Python Program to split training and testing with 80:20
# This will divide train and test , but i need to change the 4 times by changing the moldy_fruits,overripe_fruits,ripe_fruits and unripe_fruits one by one.I ran locally and uploaded
!python train_test_split.py
Traceback (most recent call last):
File "/content/gdrive/MyDrive/Colab Notebooks/hci575_data/train_test_split.py", line 21, in <module>
for subfolder in os.listdir(os.path.join(folder_path, fruit_quality)):
FileNotFoundError: [Errno 2] No such file or directory: 'C:/Users/kkumar/Downloads/HCI575/hci575_data/unripe_fruits'
pwd
'/content/gdrive/MyDrive/Colab Notebooks/hci575_data'
train_dir = "/content/gdrive/MyDrive/Colab Notebooks/hci575_data/train/"
test_dir = "/content/gdrive/MyDrive/Colab Notebooks/hci575_data/test/"
OpenCv for Image-Processing
import cv2
import os
folder_path = '/content/gdrive/MyDrive/Colab Notebooks/hci575_data/moldy_fruits/moldy_orange'
fig, ax = plt.subplots(5, 4, figsize=(10, 10))
axes = ax.flatten()
for i,filename in enumerate(os.listdir(folder_path)[:20]):
if filename.endswith('.jpg') or filename.endswith('.png'):
img = cv2.imread(os.path.join(folder_path, filename))
axes[i].imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
axes[i].axis('off')
# plt.imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
# plt.show()
#cv2.imshow('Image', img)
#cv2.waitKey(0)
#cv2.destroyAllWindows()
plt.show()
!pip install opencv-python-headless
Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Requirement already satisfied: opencv-python-headless in /usr/local/lib/python3.10/dist-packages (4.7.0.72)
Requirement already satisfied: numpy>=1.17.0 in /usr/local/lib/python3.10/dist-packages (from opencv-python-headless) (1.22.4)
cd /content/gdrive/MyDrive/Colab Notebooks/hci575_data/
/content/gdrive/MyDrive/Colab Notebooks/hci575_data
ls
apple2.jpg [0m[01;34mmoldy_fruits[0m/
[01;34mcheckpoints[0m/ orange_ripe_day.jpeg
[01;34mcheckpoints3[0m/ [01;34moverripe_fruits[0m/
fruits_classification_resnet18_model1.pt [01;34mripe_fruits[0m/
fruits_classification_resnet18_model.pt [01;34mruns[0m/
HCI575_Final_Fruit_detection.ipynb [01;34mtest[0m/
kaggle.json test.py
lime.JPG [01;34mtrain[0m/
[01;34mlogs[0m/ train_test_split.py
[01;34mlogs1[0m/ [01;34munripe_fruits[0m/
import cv2
import numpy as np
from google.colab.patches import cv2_imshow
img = cv2.imread('/content/gdrive/MyDrive/Colab Notebooks/hci575_data/apple2.jpg')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
ret, thresh = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY_INV)
lower_black = np.array([0, 0, 0])
upper_black = np.array([50,50,50])
mask = cv2.inRange(img, lower_black, upper_black)
combined_mask = cv2.bitwise_and(thresh, mask)
# contours
contours, hierarchy = cv2.findContours(combined_mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
# draw bounding boxes a
for contour in contours:
x, y, w, h = cv2.boundingRect(contour)
area = w * h
if area > 1000: # Only consider contours with area greater than 1000 pixels
cv2.rectangle(img, (x, y), (x + w, y + h), (0, 255, 0), 2)
# Display the image with the bounding boxes
cv2_imshow(img)
cv2.waitKey(0)
cv2.destroyAllWindows()
import cv2
import numpy as np
from google.colab.patches import cv2_imshow
# Load the image
img = cv2.imread('/content/gdrive/MyDrive/Colab Notebooks/hci575_data/lime.JPG')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (5, 5), 0)
circles = cv2.HoughCircles(gray, cv2.HOUGH_GRADIENT, 1, 20, param1=50, param2=30, minRadius=40, maxRadius=60)
if circles is not None:
circles = np.round(circles[0, :]).astype("int")
for (x, y, r) in circles:
cv2.circle(img, (x, y), r, (0, 255, 0), 2)
mask = np.zeros_like(gray)
cv2.circle(mask, (x, y), r, (255, 255, 255), -1)
masked_img = np.bitwise_and(gray, mask)
cropped_img = masked_img[y-r:y+r, x-r:x+r]
c = img[y-r:y+r, x-r:x+r]
ret, thresh = cv2.threshold(cropped_img, 127, 255, cv2.THRESH_BINARY_INV)
contours, hierarchy = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
#cv2.drawContours(cropped_img, contours, -1, (0, 255, 0), 2)
for contour in contours:
x, y, w, h = cv2.boundingRect(contour)
area = w * h
if area > 10 and area<5000:
color = (0, 255, 0)
cv2.drawContours(c, [contour], 0, color, 2)
# Display the image
cv2_imshow( c)
cv2.waitKey(0)
cv2.destroyAllWindows()
Training Deep Learning Model using Pretrained Model ResNet
# Packages for Deep Learning model Training
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision.transforms as transforms
import torchvision.datasets as datasets
import torchvision.models as models
from torch.utils.data import DataLoader
from torchvision.models import resnet18,resnet50
from tqdm import tqdm
from torch.utils.tensorboard import SummaryWriter
import itertools
from sklearn.metrics import classification_report, confusion_matrix
import seaborn as sns
import matplotlib.pyplot as plt
import numpy as np
# Define transforms for data augmentation and normalization
train_transform = transforms.Compose([
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
test_transform = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
# Load the dataset
train_dataset = datasets.ImageFolder('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/train', transform=train_transform)
test_dataset = datasets.ImageFolder('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/test', transform=test_transform)
# Create data loaders
train_loader = DataLoader(train_dataset, batch_size=32, shuffle=True, num_workers=2)
test_loader = DataLoader(test_dataset, batch_size=32, shuffle=False, num_workers=2)
cd /content/gdrive/MyDrive/Colab Notebooks/hci575_data/
/content/gdrive/MyDrive/Colab Notebooks/hci575_data
mkdir checkpoints
mkdir: cannot create directory ‘checkpoints’: File exists
Training Deep learning model Resnet18
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# Load ResNet18 model
#model = resnet50(num_classes=4).to(device)
model = resnet18(num_classes=4).to(device)
# Set optimizer and loss function
optimizer = optim.Adam(model.parameters())
criterion = nn.CrossEntropyLoss()
# Define the label names
label_names = ['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
def plot_confusion_matrix(cm, classes, normalize=False, title='Confusion matrix', cmap=plt.cm.Blues):
if normalize:
cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
print("Normalized confusion matrix")
else:
print('Confusion matrix, without normalization')
plt.imshow(cm, interpolation='nearest', cmap=cmap)
plt.title(title)
plt.colorbar()
tick_marks = np.arange(len(classes))
plt.xticks(tick_marks, classes, rotation=45)
plt.yticks(tick_marks, classes)
fmt = '.2f' if normalize else 'd'
thresh = cm.max() / 2.
for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):
plt.text(j, i, format(cm[i, j], fmt),
horizontalalignment="center",
color="white" if cm[i, j] > thresh else "black")
plt.tight_layout()
plt.ylabel('True label')
plt.xlabel('Predicted label')
plt.clf()
Training and Testing the ResNet18 Model
import numpy as np
label_names = ['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
writer = SummaryWriter('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/logs2')
writer.add_text('Label names', str(label_names))
num_epochs = 15
i = 0
for epoch in tqdm(range(num_epochs)):
train_loss = 0
train_acc = 0
model.train()
for images, labels in train_loader:
images = images.to(device)
labels = labels.to(device)
optimizer.zero_grad()
outputs = model(images)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
train_loss += loss.item()
train_acc += (outputs.argmax(dim=1) == labels).sum().item()
# TensorBoard
_, preds = torch.max(outputs, 1)
y_true = labels.detach().cpu().numpy()
y_pred = preds.detach().cpu().numpy()
cm = confusion_matrix(y_true, y_pred, labels=[0,1,2,3])
writer.add_image('Confusion matrix', np.expand_dims(cm, axis=2), epoch*len(train_loader) + i)
plt.clf()
i += 1
train_loss /= len(train_loader.dataset)
train_acc /= len(train_loader.dataset)
test_loss = 0
test_acc = 0
model.eval()
with torch.no_grad():
for images, labels in test_loader:
images = images.to(device)
labels = labels.to(device)
outputs = model(images)
loss = nn.functional.cross_entropy(outputs, labels)
test_loss += loss.item()
test_acc += (outputs.argmax(dim=1) == labels).sum().item()
test_loss /= len(test_loader.dataset)
test_acc /= len(test_loader.dataset)
writer.add_scalar('Training loss', train_loss/len(train_loader), epoch)
writer.add_scalar('Training accuracy', train_acc/len(train_loader.dataset), epoch)
writer.add_scalar('Testing loss', test_loss/len(test_loader), epoch)
writer.add_scalar('Testing accuracy', test_acc/len(test_loader.dataset), epoch)
print(f'Epoch {epoch + 1}/{num_epochs}, Train Loss: {train_loss:.4f}, Train Acc: {train_acc:.4f}, Test Loss: {test_loss:.4f}, Test Acc: {test_acc:.4f}')
torch.save(model.state_dict(), f'/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints2/checkpoint_epoch_{epoch+1}.pt')
# Save the trained model
torch.save(model.state_dict(), f'/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints1/fruits_classification_resnet18_model1.pt')
writer.close()
7%|▋ | 1/15 [21:13<4:57:13, 1273.81s/it]
Epoch 1/15, Train Loss: 0.0302, Train Acc: 0.6081, Test Loss: 0.0247, Test Acc: 0.6806
13%|█▎ | 2/15 [42:57<4:39:46, 1291.27s/it]
Epoch 2/15, Train Loss: 0.0246, Train Acc: 0.6878, Test Loss: 0.0229, Test Acc: 0.7038
20%|██ | 3/15 [1:09:10<4:43:58, 1419.90s/it]
Epoch 3/15, Train Loss: 0.0222, Train Acc: 0.7189, Test Loss: 0.0182, Test Acc: 0.7871
27%|██▋ | 4/15 [1:29:55<4:07:38, 1350.79s/it]
Epoch 4/15, Train Loss: 0.0211, Train Acc: 0.7343, Test Loss: 0.0163, Test Acc: 0.8224
33%|███▎ | 5/15 [1:49:52<3:35:56, 1295.60s/it]
Epoch 5/15, Train Loss: 0.0193, Train Acc: 0.7556, Test Loss: 0.0209, Test Acc: 0.7540
40%|████ | 6/15 [2:10:49<3:12:20, 1282.23s/it]
Epoch 6/15, Train Loss: 0.0183, Train Acc: 0.7702, Test Loss: 0.0132, Test Acc: 0.8428
47%|████▋ | 7/15 [2:32:43<2:52:21, 1292.70s/it]
Epoch 7/15, Train Loss: 0.0177, Train Acc: 0.7774, Test Loss: 0.0150, Test Acc: 0.8009
53%|█████▎ | 8/15 [2:53:03<2:28:06, 1269.46s/it]
Epoch 8/15, Train Loss: 0.0166, Train Acc: 0.7933, Test Loss: 0.0119, Test Acc: 0.8671
60%|██████ | 9/15 [3:15:25<2:09:13, 1292.28s/it]
Epoch 9/15, Train Loss: 0.0157, Train Acc: 0.8076, Test Loss: 0.0104, Test Acc: 0.8759
67%|██████▋ | 10/15 [3:38:48<1:50:32, 1326.50s/it]
Epoch 10/15, Train Loss: 0.0151, Train Acc: 0.8127, Test Loss: 0.0123, Test Acc: 0.8560
73%|███████▎ | 11/15 [4:08:31<1:37:44, 1466.06s/it]
Epoch 11/15, Train Loss: 0.0149, Train Acc: 0.8153, Test Loss: 0.0088, Test Acc: 0.9002
80%|████████ | 12/15 [4:30:04<1:10:40, 1413.41s/it]
Epoch 12/15, Train Loss: 0.0143, Train Acc: 0.8268, Test Loss: 0.0109, Test Acc: 0.8787
87%|████████▋ | 13/15 [4:50:02<44:56, 1348.37s/it]
Epoch 13/15, Train Loss: 0.0138, Train Acc: 0.8336, Test Loss: 0.0096, Test Acc: 0.8902
93%|█████████▎| 14/15 [5:11:05<22:02, 1322.40s/it]
Epoch 14/15, Train Loss: 0.0131, Train Acc: 0.8309, Test Loss: 0.0070, Test Acc: 0.9272
100%|██████████| 15/15 [5:55:25<00:00, 1421.69s/it]
Epoch 15/15, Train Loss: 0.0122, Train Acc: 0.8503, Test Loss: 0.0071, Test Acc: 0.9195
<Figure size 640x480 with 0 Axes>
# Load the saved ResNet50 model
model = models.resnet18(pretrained=False)
num_ftrs = model.fc.in_features
model.fc = nn.Linear(num_ftrs, 4)
model.load_state_dict(torch.load('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints1/fruits_classification_resnet18_model1.pt'))
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
test_loader = DataLoader(test_dataset, batch_size=32, shuffle=False, num_workers=2)
model.eval()
all_preds = []
all_labels = []
# Evaluate the model on the test set
with torch.no_grad():
for images, labels in test_loader:
images = images.to(device)
labels = labels.to(device)
outputs = model(images)
preds = torch.argmax(outputs, dim=1)
all_preds.extend(preds.cpu().numpy())
all_labels.extend(labels.cpu().numpy())
# confusion matrix
class_names = ['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
report = classification_report(all_labels, all_preds, target_names=class_names, digits=4)
cm = confusion_matrix(all_labels, all_preds)
cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis] # normalize the confusion matrix
fig, ax = plt.subplots(figsize=(8, 6))
sns.heatmap(cm, annot=True, cmap='Blues', fmt='.2f', xticklabels=class_names, yticklabels=class_names, ax=ax)
ax.set_xlabel('Predicted label')
ax.set_ylabel('True label')
ax.set_title('Confusion Matrix')
plt.tight_layout()
/Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.
warnings.warn(
/Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=None`.
warnings.warn(msg)
import torch
import numpy as np
from sklearn.metrics import confusion_matrix, precision_score, recall_score, f1_score
# Load the saved model
model_path = '/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints1/fruits_classification_resnet18_model1.pt'
model_state_dict = torch.load(model_path)
model = models.resnet18(pretrained=True)
num_classes = 4
num_ftrs = model.fc.in_features
model.fc = nn.Linear(num_ftrs, num_classes)
model.load_state_dict(model_state_dict)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
model.eval()
# Define the function to predict labels
def predict(model, dataloader):
predictions = []
true_labels = []
with torch.no_grad():
for inputs, labels in dataloader:
inputs = inputs.to(device)
labels = labels.to(device)
outputs = model(inputs)
_, predicted = torch.max(outputs.data, 1)
predictions.extend(predicted.cpu().numpy())
true_labels.extend(labels.cpu().numpy())
return predictions, true_labels
# Get the predictions and true labels for the test set
test_preds, test_labels = predict(model, test_loader)
# Calculate the confusion matrix, precision, recall, and F1 score
cm = confusion_matrix(test_labels, test_preds)
precision = precision_score(test_labels, test_preds, average=None)
recall = recall_score(test_labels, test_preds, average=None)
f1 = f1_score(test_labels, test_preds, average=None)
# Print the precision, recall, and F1 score for each class
label_names = ['Moldy Fruits', 'Overripe Fruits', 'Ripe Fruits', 'Unripe Fruits']
for i, label in enumerate(label_names):
print(f"{label} - Precision: {precision[i]:.4f}, Recall: {recall[i]:.4f}, F1 score: {f1[i]:.4f}")
# Print the overall precision, recall, and F1 score
print(f"Overall - Precision: {np.mean(precision):.4f}, Recall: {np.mean(recall):.4f}, F1 score: {np.mean(f1):.4f}")
/Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.
warnings.warn(
/Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ResNet18_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet18_Weights.DEFAULT` to get the most up-to-date weights.
warnings.warn(msg)
Moldy Fruits - Precision: 0.8709, Recall: 0.9031, F1 score: 0.8867
Overripe Fruits - Precision: 0.9266, Recall: 0.9011, F1 score: 0.9136
Ripe Fruits - Precision: 0.9356, Recall: 0.9423, F1 score: 0.9390
Unripe Fruits - Precision: 0.9331, Recall: 0.9280, F1 score: 0.9306
Overall - Precision: 0.9165, Recall: 0.9186, F1 score: 0.9175
Training Deep learning model Resnet50
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# Load ResNet50 model
model = resnet50(num_classes=4).to(device)
#model = resnet18(num_classes=4).to(device)
# Set optimizer and loss function
optimizer = optim.Adam(model.parameters())
criterion = nn.CrossEntropyLoss()
label_names = ['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
writer = SummaryWriter('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/logs1')
writer.add_text('Label names', str(label_names))
num_epochs = 15
i = 0
for epoch in tqdm(range(num_epochs)):
train_loss = 0
train_acc = 0
model.train()
for images, labels in train_loader:
images = images.to(device)
labels = labels.to(device)
optimizer.zero_grad()
outputs = model(images)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
train_loss += loss.item()
train_acc += (outputs.argmax(dim=1) == labels).sum().item()
# TensorBoard
_, preds = torch.max(outputs, 1)
y_true = labels.detach().cpu().numpy()
y_pred = preds.detach().cpu().numpy()
cm = confusion_matrix(y_true, y_pred, labels=[0,1,2,3])
writer.add_image('Confusion matrix', np.expand_dims(cm, axis=2), epoch*len(train_loader) + i)
plt.clf()
i += 1
train_loss /= len(train_loader.dataset)
train_acc /= len(train_loader.dataset)
test_loss = 0
test_acc = 0
model.eval()
with torch.no_grad():
for images, labels in test_loader:
images = images.to(device)
labels = labels.to(device)
outputs = model(images)
loss = nn.functional.cross_entropy(outputs, labels)
test_loss += loss.item()
test_acc += (outputs.argmax(dim=1) == labels).sum().item()
test_loss /= len(test_loader.dataset)
test_acc /= len(test_loader.dataset)
writer.add_scalar('Training loss', train_loss/len(train_loader), epoch)
writer.add_scalar('Training accuracy', train_acc/len(train_loader.dataset), epoch)
writer.add_scalar('Testing loss', test_loss/len(test_loader), epoch)
writer.add_scalar('Testing accuracy', test_acc/len(test_loader.dataset), epoch)
print(f'Epoch {epoch + 1}/{num_epochs}, Train Loss: {train_loss:.4f}, Train Acc: {train_acc:.4f}, Test Loss: {test_loss:.4f}, Test Acc: {test_acc:.4f}')
torch.save(model.state_dict(), f'/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints2/checkpoint_epoch_{epoch+1}.pt')
# Save the trained model
torch.save(model.state_dict(), f'/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints1/fruits_classification_resnet18_model1.pt')
writer.close()
ResNet50 Model Loading and confusion Matrix
import torch
from torch.utils.data import DataLoader
from sklearn.metrics import classification_report, confusion_matrix
import seaborn as sns
import matplotlib.pyplot as plt
import numpy as np
# Load the saved ResNet50 model
model = models.resnet50(pretrained=False)
num_ftrs = model.fc.in_features
model.fc = nn.Linear(num_ftrs, 4)
model.load_state_dict(torch.load('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints1/fruits_classification_resnet50_model.pt'))
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
test_loader = DataLoader(test_dataset, batch_size=32, shuffle=False, num_workers=2)
model.eval()
all_preds = []
all_labels = []
# Evaluation
with torch.no_grad():
for images, labels in test_loader:
images = images.to(device)
labels = labels.to(device)
outputs = model(images)
preds = torch.argmax(outputs, dim=1)
all_preds.extend(preds.cpu().numpy())
all_labels.extend(labels.cpu().numpy())
# confusion matrix
class_names = ['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
report = classification_report(all_labels, all_preds, target_names=class_names, digits=4)
cm = confusion_matrix(all_labels, all_preds)
cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis] # normalize the confusion matrix
fig, ax = plt.subplots(figsize=(8, 6))
sns.heatmap(cm, annot=True, cmap='Blues', fmt='.2f', xticklabels=class_names, yticklabels=class_names, ax=ax)
ax.set_xlabel('Predicted label')
ax.set_ylabel('True label')
ax.set_title('Confusion Matrix')
plt.tight_layout()
/Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.
warnings.warn(
/Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=None`.
warnings.warn(msg)
from sklearn.metrics import precision_score, recall_score
# Calculate precision and recall for each label
precision = precision_score(all_labels, all_preds, average=None)
recall = recall_score(all_labels, all_preds, average=None)
# Print precision and recall for each label
for i, class_name in enumerate(class_names):
print(f'{class_name} - Precision: {precision[i]:.4f}, Recall: {recall[i]:.4f}')
moldy_fruits - Precision: 0.6667, Recall: 0.7464
overripe_fruits - Precision: 0.8324, Recall: 0.8004
ripe_fruits - Precision: 0.7702, Recall: 0.4468
unripe_fruits - Precision: 0.5724, Recall: 0.9086
from sklearn.metrics import confusion_matrix, precision_score, recall_score, f1_score
# Load the saved model
model_path = '/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints1/fruits_classification_resnet50_model.pt'
model_state_dict = torch.load(model_path)
model = models.resnet50(pretrained=True)
num_classes = 4
num_ftrs = model.fc.in_features
model.fc = nn.Linear(num_ftrs, num_classes)
model.load_state_dict(model_state_dict)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
model.eval()
# Define the function to predict labels
def predict(model, dataloader):
predictions = []
true_labels = []
with torch.no_grad():
for inputs, labels in dataloader:
inputs = inputs.to(device)
labels = labels.to(device)
outputs = model(inputs)
_, predicted = torch.max(outputs.data, 1)
predictions.extend(predicted.cpu().numpy())
true_labels.extend(labels.cpu().numpy())
return predictions, true_labels
test_preds, test_labels = predict(model, test_loader)
cm = confusion_matrix(test_labels, test_preds)
precision = precision_score(test_labels, test_preds, average=None)
recall = recall_score(test_labels, test_preds, average=None)
f1 = f1_score(test_labels, test_preds, average=None)
test_preds = torch.tensor(test_preds)
test_labels = torch.tensor(test_labels)
accuracy = (test_preds == test_labels).float().mean()
# precision, recall, F1 score, and accuracy for each class
label_names = ['Moldy Fruits', 'Overripe Fruits', 'Ripe Fruits', 'Unripe Fruits']
for i, label in enumerate(label_names):
print(f"{label} - Precision: {precision[i]:.4f}, Recall: {recall[i]:.4f}, F1 score: {f1[i]:.4f}")
print(f"Accuracy: {accuracy:.4f}")
# overall precision, recall, F1 score, and accuracy
print(f"Overall - Precision: {np.mean(precision):.4f}, Recall: {np.mean(recall):.4f}, F1 score: {np.mean(f1):.4f}, Accuracy: {accuracy:.4f}")
/Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.
warnings.warn(
/Users/kundankumar/opt/anaconda3/envs/hci575_data/lib/python3.8/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ResNet50_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet50_Weights.DEFAULT` to get the most up-to-date weights.
warnings.warn(msg)
Moldy Fruits - Precision: 0.6667, Recall: 0.7464, F1 score: 0.7043
Overripe Fruits - Precision: 0.8324, Recall: 0.8004, F1 score: 0.8161
Ripe Fruits - Precision: 0.7702, Recall: 0.4468, F1 score: 0.5656
Unripe Fruits - Precision: 0.5724, Recall: 0.9086, F1 score: 0.7024
Accuracy: 0.7033
Overall - Precision: 0.7104, Recall: 0.7256, F1 score: 0.6971, Accuracy: 0.7033
print(test_dataset.classes)
['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
Convoltional Neural Network training from Scratch
label_names = ['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
writer = SummaryWriter('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/logs3')
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.fc1 = nn.Linear(224*224*3, 256)
self.fc2 = nn.Linear(256, 128)
self.fc3 = nn.Linear(128, 4)
def forward(self, x):
x = x.view(-1, 224*224*3)
x = nn.functional.relu(self.fc1(x))
x = nn.functional.relu(self.fc2(x))
x = self.fc3(x)
return x
model = Net()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
# Define the loss function and optimizer
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)
# Train the model
num_epochs = 15
for epoch in range(num_epochs):
train_loss = 0
train_acc = 0
model.train()
for images, labels in train_loader:
images = images.to(device)
labels = labels.to(device)
optimizer.zero_grad()
outputs = model(images)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
train_loss += loss.item()
train_acc += (outputs.argmax(dim=1) == labels).sum().item()
train_loss /= len(train_loader.dataset)
train_acc /= len(train_loader.dataset)
writer.add_scalar('Loss/train', train_loss, epoch)
writer.add_scalar('Accuracy/train', train_acc, epoch)
writer.add_scalar('Loss/test', test_loss, epoch)
writer.add_scalar('Accuracy/test', test_acc, epoch)
test_loss = 0
test_acc = 0
model.eval()
with torch.no_grad():
for images, labels in test_loader:
images = images.to(device)
labels = labels.to(device)
outputs = model(images)
loss = criterion(outputs, labels)
test_loss += loss.item()
test_acc += (outputs.argmax(dim=1) == labels).sum().item()
test_loss /= len(test_loader.dataset)
test_acc /= len(test_loader.dataset)
print(f'Epoch {epoch + 1}/{num_epochs}, Train Loss: {train_loss:.4f}, Train Acc: {train_acc:.4f}, Test Loss: {test_loss:.4f}, Test Acc: {test_acc:.4f}')
torch.save(model.state_dict(), f'/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints3/checkpoint_epoch_{epoch+1}.pt')
# Save the trained model
#torch.save(model.state_dict(), '/content/gdrive/MyDrive/Colab Notebooks/hci575_data/my_model.pt')
torch.save(model.state_dict(), f'/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints3/fruits_classification_scratch_model.pt')
Epoch 1/15, Train Loss: 0.0350, Train Acc: 0.4903, Test Loss: 0.0285, Test Acc: 0.5880
Epoch 2/15, Train Loss: 0.0304, Train Acc: 0.5825, Test Loss: 0.0245, Test Acc: 0.6586
Epoch 3/15, Train Loss: 0.0290, Train Acc: 0.6000, Test Loss: 0.0218, Test Acc: 0.7270
Epoch 4/15, Train Loss: 0.0275, Train Acc: 0.6230, Test Loss: 0.0204, Test Acc: 0.7391
Epoch 5/15, Train Loss: 0.0265, Train Acc: 0.6372, Test Loss: 0.0190, Test Acc: 0.7672
Epoch 6/15, Train Loss: 0.0258, Train Acc: 0.6528, Test Loss: 0.0180, Test Acc: 0.7634
Epoch 7/15, Train Loss: 0.0253, Train Acc: 0.6573, Test Loss: 0.0164, Test Acc: 0.8064
Epoch 8/15, Train Loss: 0.0244, Train Acc: 0.6743, Test Loss: 0.0175, Test Acc: 0.7678
Epoch 9/15, Train Loss: 0.0242, Train Acc: 0.6758, Test Loss: 0.0150, Test Acc: 0.8229
Epoch 10/15, Train Loss: 0.0235, Train Acc: 0.6921, Test Loss: 0.0154, Test Acc: 0.8257
Epoch 11/15, Train Loss: 0.0234, Train Acc: 0.6930, Test Loss: 0.0138, Test Acc: 0.8323
Epoch 12/15, Train Loss: 0.0226, Train Acc: 0.7046, Test Loss: 0.0142, Test Acc: 0.8356
Epoch 13/15, Train Loss: 0.0225, Train Acc: 0.7015, Test Loss: 0.0130, Test Acc: 0.8478
Epoch 14/15, Train Loss: 0.0218, Train Acc: 0.7134, Test Loss: 0.0127, Test Acc: 0.8577
Epoch 15/15, Train Loss: 0.0214, Train Acc: 0.7272, Test Loss: 0.0116, Test Acc: 0.8731
!pip install torchsummary
import torch
from torchsummary import summary
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.fc1 = nn.Linear(224*224*3, 256)
self.fc2 = nn.Linear(256, 128)
self.fc3 = nn.Linear(128, 4)
def forward(self, x):
x = x.view(-1, 224*224*3)
x = nn.functional.relu(self.fc1(x))
x = nn.functional.relu(self.fc2(x))
x = self.fc3(x)
return x
model = Net()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
summary(model, input_size=(3, 224, 224))
Collecting torchsummary
Downloading torchsummary-1.5.1-py3-none-any.whl (2.8 kB)
Installing collected packages: torchsummary
Successfully installed torchsummary-1.5.1
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Linear-1 [-1, 256] 38,535,424
Linear-2 [-1, 128] 32,896
Linear-3 [-1, 4] 516
================================================================
Total params: 38,568,836
Trainable params: 38,568,836
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.57
Forward/backward pass size (MB): 0.00
Params size (MB): 147.13
Estimated Total Size (MB): 147.71
----------------------------------------------------------------
# Define the neural network model
class Net(torch.nn.Module):
def __init__(self):
super(Net, self).__init__()
self.fc1 = torch.nn.Linear(224*224*3, 256)
self.fc2 = torch.nn.Linear(256, 128)
self.fc3 = torch.nn.Linear(128, 4)
def forward(self, x):
x = x.view(-1, 224*224*3)
x = torch.nn.functional.relu(self.fc1(x))
x = torch.nn.functional.relu(self.fc2(x))
x = self.fc3(x)
return x
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
import torch
import numpy as np
from sklearn.metrics import confusion_matrix
import seaborn as sns
import matplotlib.pyplot as plt
label_names = ['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
model = Net()
model.load_state_dict(torch.load('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints3/fruits_classification_scratch_model.pt'))
model.eval()
# Create a function to predict labels
def predict(model, dataloader):
predictions = []
true_labels = []
with torch.no_grad():
for inputs, labels in dataloader:
inputs = inputs.to(device)
labels = labels.to(device)
outputs = model(inputs)
_, predicted = torch.max(outputs.data, 1)
predictions.extend(predicted.cpu().numpy())
true_labels.extend(labels.cpu().numpy())
return predictions, true_labels
# Get the predictions and true labels for the test set
test_preds, test_labels = predict(model, test_loader)
cm = confusion_matrix(test_labels, test_preds, normalize='true')
# Plot the confusion matrix
sns.set(font_scale=1.4)
sns.heatmap(cm, annot=True, cmap='Blues', xticklabels=label_names, yticklabels=label_names, fmt='.2f')
plt.xlabel('Predicted')
plt.ylabel('True')
plt.show()
CNN Model Precision and Accuracy
model = Net()
model.load_state_dict(torch.load('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints3/fruits_classification_scratch_model.pt'))
model.eval()
def predict(model, dataloader):
predictions = []
true_labels = []
with torch.no_grad():
for inputs, labels in dataloader:
inputs = inputs.to(device)
labels = labels.to(device)
outputs = model(inputs)
_, predicted = torch.max(outputs.data, 1)
predictions.extend(predicted.cpu().numpy())
true_labels.extend(labels.cpu().numpy())
return predictions, true_labels
# Get the predictions and true labels for the test set
test_preds, test_labels = predict(model, test_loader)
cm = confusion_matrix(test_labels, test_preds)
tp = np.diag(cm)
fp = np.sum(cm, axis=0) - tp
fn = np.sum(cm, axis=1) - tp
precision = tp / (tp + fp)
recall = tp / (tp + fn)
f1 = 2 * (precision * recall) / (precision + recall)
for i, label in enumerate(label_names):
print(f"{label} - Precision: {precision[i]:.4f}, Recall: {recall[i]:.4f}, F1 score: {f1[i]:.4f}")
moldy_fruits - Precision: 0.8713, Recall: 0.8291, F1 score: 0.8496
overripe_fruits - Precision: 0.8675, Recall: 0.8993, F1 score: 0.8831
ripe_fruits - Precision: 0.8494, Recall: 0.9045, F1 score: 0.8761
unripe_fruits - Precision: 0.9286, Recall: 0.8283, F1 score: 0.8755
model = Net()
model.load_state_dict(torch.load('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints3/fruits_classification_scratch_model.pt'))
model.eval()
# Define the loss function and optimizer
criterion = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters())
train_loss_list = []
test_loss_list = []
train_acc_list = []
test_acc_list = []
# Set up a function to calculate the accuracy
def calculate_accuracy(predictions, labels):
correct = np.sum(predictions == labels)
total = len(predictions)
accuracy = correct / total
return accuracy
num_epochs = 15
for epoch in range(num_epochs):
# Train the model
model.train()
train_loss = 0.0
train_correct = 0
train_total = 0
for inputs, labels in train_loader:
optimizer.zero_grad()
outputs = model(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
train_loss += loss.item() * inputs.size(0)
_, predicted = torch.max(outputs.data, 1)
train_correct += (predicted == labels).sum().item()
train_total += labels.size(0)
train_loss = train_loss / len(train_dataset)
train_accuracy = calculate_accuracy(predicted.cpu().numpy(), labels.cpu().numpy())
model.eval()
test_loss = 0.0
test_correct = 0
test_total = 0
with torch.no_grad():
for inputs, labels in test_loader:
outputs = model(inputs)
loss = criterion(outputs, labels)
test_loss += loss.item() * inputs.size(0)
_, predicted = torch.max(outputs.data, 1)
test_correct += (predicted == labels).sum().item()
test_total += labels.size(0)
test_loss = test_loss / len(test_dataset)
test_accuracy = calculate_accuracy(predicted.cpu().numpy(), labels.cpu().numpy())
# Append the loss and accuracy values to the lists
train_loss_list.append(train_loss)
test_loss_list.append(test_loss)
train_acc_list.append(train_accuracy)
test_acc_list.append(test_accuracy)
# Plot the loss and accuracy values for each epoch
x = range(num_epochs)
plt.plot(x, train_loss_list, label='Training Loss')
plt.plot(x, test_loss_list, label='Testing Loss')
plt.plot(x, train_acc_list, label='Training Accuracy')
plt.plot(x, test_acc_list, label='Testing Accuracy')
plt.legend(loc='best')
plt.xlabel('Epoch')
plt.ylabel('Value')
plt.title('Training and Testing Loss and Accuracy')
plt.show()
Testing the Model
print(test_dataset.classes)
['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
import torch
from PIL import Image
from torchvision import transforms
# Load the saved ResNet50 model
model = models.resnet50(pretrained=False)
num_ftrs = model.fc.in_features
model.fc = nn.Linear(num_ftrs, 4)
model.load_state_dict(torch.load('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints1/fruits_classification_resnet50_model.pt'))
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
preprocess = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
# Load and preprocess the image
image_path = '/Users/kundankumar/Downloads/evening_dim_lighting/IMG_9886.jpeg'
image = Image.open(image_path)
image_tensor = preprocess(image)
image_batch = torch.unsqueeze(image_tensor, 0)
image_batch = image_batch.to(device)
# Set the model to evaluation mode
model.eval()
# Feed the image through the model and get the predicted class
with torch.no_grad():
outputs = model(image_batch)
_, preds = torch.max(outputs, 1)
predicted_class = preds[0].item()
class_names = ['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
print('Predicted class: ', class_names[predicted_class])
Predicted class: ripe_fruits
import torch
from PIL import Image
from torchvision import transforms
import matplotlib.pyplot as plt
# Load the saved ResNet50 model
model = models.resnet18(pretrained=False)
num_ftrs = model.fc.in_features
model.fc = nn.Linear(num_ftrs, 4)
model.load_state_dict(torch.load('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints1/fruits_classification_resnet18_model1.pt'))
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
preprocess = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
class_names = ['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
# Load and preprocess the image
image_path = '/Users/kundankumar/Downloads/Sorted test images/banana_overripe_day_flash.jpeg'
image = Image.open(image_path)
with torch.no_grad():
image_tensor = preprocess(image)
image_batch = torch.unsqueeze(image_tensor, 0)
image_batch = image_batch.to(device)
outputs = model(image_batch)
_, preds = torch.max(outputs, 1)
predicted_class = preds[0].item()
# boung_box
linewidth = 5
bbox = {'facecolor': 'white', 'alpha': 0.7, 'pad': 10}
fig, ax = plt.subplots(figsize=(8, 8))
ax.imshow(image)
text = 'Predicted class: ' + class_names[predicted_class]
ax.text(0.5, 0.95, text, transform=ax.transAxes, fontsize=24,
verticalalignment='top', bbox=bbox)
rect = plt.Rectangle((0, 0), image.width, image.height, linewidth=linewidth,
edgecolor='r', facecolor='none')
ax.add_patch(rect)
ax.set_xticks([])
ax.set_yticks([])
ax.set_xlabel('')
ax.set_ylabel('')
plt.show()
import torch
from PIL import Image
from torchvision import transforms
import os
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.patches as patches
model = models.resnet50(pretrained=False)
num_ftrs = model.fc.in_features
model.fc = nn.Linear(num_ftrs, 4)
model.load_state_dict(torch.load('/Users/kundankumar/Documents/COMS_575/Project/hci575_data/checkpoints1/fruits_classification_resnet50_model.pt'))
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
preprocess = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
class_names = ['moldy_fruits', 'overripe_fruits', 'ripe_fruits', 'unripe_fruits']
fig, axs = plt.subplots(1, len(os.listdir('/Users/kundankumar/Downloads/Sorted test images/')), figsize=(15, 5))
for i, image_path in enumerate(os.listdir('/Users/kundankumar/Downloads/Sorted test images/')):
image = Image.open(os.path.join('/Users/kundankumar/Downloads/Sorted test images/', image_path))
image_tensor = preprocess(image)
image_batch = torch.unsqueeze(image_tensor, 0)
image_batch = image_batch.to(device)
model.eval()
with torch.no_grad():
outputs = model(image_batch)
_, preds = torch.max(outputs, 1)
predicted_class = preds[0].item()
ax = axs[i]
ax.imshow(image)
ax.axis('off')
ax.set_title(class_names[predicted_class])
box = patches.Rectangle((10, 10), 200, 200, linewidth=2, edgecolor='r', facecolor='none')
ax.add_patch(box)
# Display the figure
plt.tight_layout()
plt.show()
