| Website | \nLinks | \n
|---|---|
| GitHub | \n![\"GitHub\"/](\"https://img.shields.io/badge/GitHub-Profile-blue?style=for-the-badge&logo=github\"){kind=icon} | \n
![\"LinkedIn\"/](\"https://img.shields.io/badge/LinkedIn-Profile-blue?style=for-the-badge&logo=linkedin\"){kind=icon} | \n |
![\"Facebook\"/](\"https://img.shields.io/badge/Facebook-Profile-blue?style=for-the-badge&logo=facebook\"){kind=icon} | \n |
| Gmail | \n![\"Gmail\"/](\"https://img.shields.io/badge/Gmail-Contact%20Me-red?style=for-the-badge&logo=gmail\"){kind=icon} | \n
Paddy Doctor: Paddy Disease Classification | 97% Train | 95% Test Accuracy
\n\n\n##### **About Author!
\n##### ***Hello, I'm Muhammad_Abdullah : A Data Science Enthusiast and Kaggle 2x Expert***\n\nGreetings! I'm delighted to welcome you into my world of data science exploration and innovation. I'm **Muhammad_Abdullah**, a passionate data scientist with a fervent dedication to unraveling the mysteries hidden within datasets and leveraging the power of machine learning to drive meaningful insights and solutions.\n\n###### ***A Passion for Data Science***\n\nSince the inception of my journey into the captivating realm of data science, I've been driven by an insatiable curiosity and an unwavering passion for uncovering the stories embedded in data. From the thrill of diving deep into complex datasets to the exhilaration of crafting predictive models that shape our understanding of the world, data science has become not just a profession but a lifelong passion.\n\n###### ***Guiding Light on Kaggle***\n\nAs a Kaggle 2x Expert, I've had the privilege of sharing my knowledge, insights, and experiences with the vibrant Kaggle community. Through meticulously crafted notebooks, engaging discussions, and collaborative projects, I've had the opportunity to mentor aspiring data enthusiasts, foster a culture of learning and growth, and contribute to the collective pursuit of excellence in data science.\n\n###### ***Let's Explore Together***\n\nJoin me on an exhilarating adventure into the dynamic world of data science! Together, we'll unlock the potential of data, unravel its mysteries, and embark on a transformative journey of discovery and innovation. Whether you're a seasoned data enthusiast or just beginning your data science odyssey, I'm excited to share this journey with you and explore the endless possibilities that data science has to offer.\n| Website | \nLinks | \n
|---|---|
| GitHub | \n | \n
| \n |
| \n |
| Gmail | \n | \n
Importing Libraries
","metadata":{}},{"cell_type":"code","source":"#Import Os and Basis Libraries\nimport cv2\nimport os\nimport pandas as pd\nimport numpy as np\nimport seaborn as sns\nimport matplotlib.pyplot as plt \n#Matplot Images\nimport matplotlib.image as mpimg\n# Tensflor and Keras Layer and Model and Optimize and Loss\nimport tensorflow as tf\nfrom tensorflow import keras\nfrom keras import Sequential\nfrom keras.layers import *\nfrom tensorflow.keras.losses import BinaryCrossentropy\n#PreTrained Model\nfrom tensorflow.keras.applications import *\n#Image Generator DataAugmentation\n#Early Stopping\nfrom tensorflow.keras.callbacks import EarlyStopping\n# Warnings Remove \nimport warnings \nwarnings.filterwarnings(\"ignore\")\n#Splitting Data \n# import splitfolders\nfrom tensorflow.keras.preprocessing import image_dataset_from_directory\nimport random\nfrom sklearn.cluster import KMeans\nfrom tensorflow.keras.applications.resnet50 import preprocess_input\nfrom tensorflow.keras.preprocessing import image","metadata":{"execution":{"iopub.execute_input":"2024-02-25T18:36:13.677477Z","iopub.status.busy":"2024-02-25T18:36:13.677136Z","iopub.status.idle":"2024-02-25T18:36:28.455798Z","shell.execute_reply":"2024-02-25T18:36:28.454872Z","shell.execute_reply.started":"2024-02-25T18:36:13.677448Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"#Loading Data
\n","metadata":{}},{"cell_type":"code","source":"# Directory containing the \"Train\" folder\ndirectory = \"/kaggle/input/paddy-disease-classification/train_images\"\n\nfilepath =[]\nlabel = []\n\nfolds = os.listdir(directory)\n\nfor fold in folds:\n f_path = os.path.join(directory , fold)\n \n imgs = os.listdir(f_path)\n \n for img in imgs:\n \n img_path = os.path.join(f_path , img)\n filepath.append(img_path)\n label.append(fold)\n \n#Concat data paths with labels\nfile_path_series = pd.Series(filepath , name= 'filepath')\nLabel_path_series = pd.Series(label , name = 'label')\ndf_train = pd.concat([file_path_series ,Label_path_series ] , axis = 1)","metadata":{"execution":{"iopub.execute_input":"2024-02-25T18:38:01.618892Z","iopub.status.busy":"2024-02-25T18:38:01.618123Z","iopub.status.idle":"2024-02-25T18:38:03.920984Z","shell.execute_reply":"2024-02-25T18:38:03.920172Z","shell.execute_reply.started":"2024-02-25T18:38:01.618860Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Head \ndf_train","metadata":{"execution":{"iopub.execute_input":"2024-02-25T18:38:03.922887Z","iopub.status.busy":"2024-02-25T18:38:03.922574Z","iopub.status.idle":"2024-02-25T18:38:03.941597Z","shell.execute_reply":"2024-02-25T18:38:03.940700Z","shell.execute_reply.started":"2024-02-25T18:38:03.922857Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"#Train Test Val Datasets
\n","metadata":{}},{"cell_type":"code","source":"# Spliting in Train and Val \ntrain_df , val_df = train_test_split(df_train ,train_size = 0.9 , shuffle = True ,random_state = 42 )","metadata":{"execution":{"iopub.execute_input":"2024-02-25T18:38:03.943914Z","iopub.status.busy":"2024-02-25T18:38:03.942937Z","iopub.status.idle":"2024-02-25T18:38:03.951260Z","shell.execute_reply":"2024-02-25T18:38:03.950425Z","shell.execute_reply.started":"2024-02-25T18:38:03.943882Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Shape\nprint(f\"The shape of The Train data is: {train_df.shape}\")\nprint(f\"The shape of The Validation data is: {val_df.shape}\")","metadata":{"execution":{"iopub.execute_input":"2024-02-25T18:38:03.953889Z","iopub.status.busy":"2024-02-25T18:38:03.953548Z","iopub.status.idle":"2024-02-25T18:38:03.961901Z","shell.execute_reply":"2024-02-25T18:38:03.960965Z","shell.execute_reply.started":"2024-02-25T18:38:03.953857Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"###### **Visualizing Some Images From Data
\n","metadata":{}},{"cell_type":"markdown","source":"###### **Model Building
\n","metadata":{}},{"cell_type":"markdown","source":"###### **Predictions On Unseen Data
\n","metadata":{}},{"cell_type":"markdown","source":"> ## ***I Use The Following Apporch to Do Predictions on Unseen and Un Labled Data.***\n\n### ***Image Clustering and Labeling with K-Means***\n\n#### ***Introduction:***\n***This code snippet demonstrates the process of clustering and labeling images using the K-Means algorithm based on their extracted features. The images are sourced from a directory containing unlabeled images.***\n\n#### ***Steps:***\n\n1. **Importing Libraries:** ***The necessary libraries including NumPy, scikit-learn, TensorFlow, and Matplotlib are imported.***\n\n2. **Setting Directory and Class Names:** ***The directory containing unlabeled images is defined, and a dictionary mapping predicted class indices to class names is created.***\n3. **Random Image Selection:** ***20 image files are randomly selected from the directory for processing.***\n\n4. **Feature Extraction:** ***Features are extracted from each image using the ResNet50 model. The images are preprocessed before feature extraction.***\n\n5. **K-Means Clustering:** ***The extracted features are used to perform K-Means clustering with 10 clusters.***\n\n6. **Displaying Images with Predicted Labels:** ***The images are displayed along with their predicted labels. Predicted labels are determined by mapping predicted class indices to class names.***\n\n#### ***Functionality:***\n\n- **display_images_with_labels:** ***This function displays the images with their predicted labels.***\n- **Predicted Labels:** ***Predicted labels for each image are determined using the trained model and mapped to class names.***\n\n#### ***Result:***\n***The code successfully clusters and labels the images based on their features, providing insights into the content of the images and facilitating further analysis or classification tasks.***\n","metadata":{}},{"cell_type":"code","source":"\n\n# Define the directory containing unlabeled images\nimages_dir = '/kaggle/input/paddy-disease-classification/test_images'\n\n# Define a dictionary to map predicted class indices to class names\nclass_names = {\n 0: 'bacterial_leaf_blight',\n 1: 'bacterial_leaf_streak',\n 2: 'bacterial_panicle_blight',\n 3: 'blast',\n 4: 'brown_spot',\n 5: 'dead_heart',\n 6: 'downy_mildew',\n 7: 'hispa',\n 8: 'normal',\n 9: 'tungro'\n}\n\n# Randomly select 20 image files from the directory\nimage_files = os.listdir(images_dir)\nrandom_images = random.sample(image_files, 20)\n\n\n# Extract features from each image\nfeatures = []\nfor img_file in random_images:\n img_path = os.path.join(images_dir, img_file)\n img = image.load_img(img_path, target_size=(224, 224))\n x = image.img_to_array(img)\n x = np.expand_dims(x, axis=0)\n x = preprocess_input(x)\n features.append(model.predict(x).flatten())\n\n# Convert list of feature vectors to numpy array\nfeatures = np.array(features)\n\n# Perform K-means clustering\nkmeans = KMeans(n_clusters=10, random_state=42)\nclusters = kmeans.fit_predict(features)\n\n# Function to display images with predicted labels\ndef display_images_with_labels(images, labels):\n plt.figure(figsize=(15, 10))\n for i, img_file in enumerate(images):\n img_path = os.path.join(images_dir, img_file)\n img = image.load_img(img_path, target_size=(224, 224))\n plt.subplot(4, 5, i + 1)\n plt.imshow(img)\n plt.title(f\"Predicted Label: {labels[i]}\")\n plt.axis('off')\n plt.show()\n\n# Predicted labels for each image\npredicted_labels = []\n\n# Print predicted labels for each image and display images\nfor img_file in random_images:\n img_path = os.path.join(images_dir, img_file)\n img = image.load_img(img_path, target_size=(224, 224))\n x = image.img_to_array(img)\n x = np.expand_dims(x, axis=0)\n x = preprocess_input(x)\n predicted_label_index = model.predict(x)\n predicted_label = class_names[np.argmax(predicted_label_index)] # Map predicted class index to class name\n predicted_labels.append(predicted_label)\n\n# Display images with predicted labels\ndisplay_images_with_labels(random_images, predicted_labels)\n","metadata":{"execution":{"iopub.execute_input":"2024-02-25T20:01:58.389704Z","iopub.status.busy":"2024-02-25T20:01:58.388822Z","iopub.status.idle":"2024-02-25T20:02:03.781191Z","shell.execute_reply":"2024-02-25T20:02:03.780165Z","shell.execute_reply.started":"2024-02-25T20:01:58.389669Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"***Well These are the Models Predictions , I have used Clustering methods to Do Predictions , Beacuse train Data is Not Labeled. Well Our Model is Achiveing a Good Accuracy of Testing 95% Which is Good . Hope u find Notebook Usefull. Do UpVote and Comment If You Find Notebook UseFull.***","metadata":{}},{"cell_type":"markdown","source":"#The End
\n","metadata":{}}]}