diff --git a/Acne_Prediction_Model.ipynb b/Acne_Prediction_Model.ipynb new file mode 100644 index 00000000..4ad23167 --- /dev/null +++ b/Acne_Prediction_Model.ipynb @@ -0,0 +1 @@ +{"metadata":{"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[{"sourceId":7547567,"sourceType":"datasetVersion","datasetId":4395695},{"sourceId":6100,"sourceType":"modelInstanceVersion","modelInstanceId":4643}],"dockerImageVersionId":30646,"isInternetEnabled":false,"language":"python","sourceType":"notebook","isGpuEnabled":true},"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.10.13"},"papermill":{"default_parameters":{},"duration":2123.85405,"end_time":"2024-02-04T00:06:48.395698","environment_variables":{},"exception":true,"input_path":"__notebook__.ipynb","output_path":"__notebook__.ipynb","parameters":{},"start_time":"2024-02-03T23:31:24.541648","version":"2.5.0"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"# Importing dependencies\n\nimport os\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport cv2\nfrom tqdm.notebook import tqdm\n\nimport tensorflow as tf\nfrom tensorflow.keras import *\nfrom tensorflow.keras.optimizers import AdamW\nfrom tensorflow.keras.callbacks import *\nimport keras_cv\n\n\nBATCH_SIZE = 16\nAUTO = tf.data.AUTOTUNE","metadata":{"_kg_hide-output":true,"papermill":{"duration":19.027459,"end_time":"2024-02-03T23:31:46.273022","exception":false,"start_time":"2024-02-03T23:31:27.245563","status":"completed"},"tags":[],"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Preprocessing","metadata":{"papermill":{"duration":0.005875,"end_time":"2024-02-03T23:31:46.285376","exception":false,"start_time":"2024-02-03T23:31:46.279501","status":"completed"},"tags":[]}},{"cell_type":"code","source":"# a function for converting txt file to list\ndef parse_txt_annot(img_path, txt_path):\n img = cv2.imread(img_path)\n w = int(img.shape[0])\n h = int(img.shape[1])\n\n file_label = open(txt_path, \"r\")\n lines = file_label.read().split('\\n')\n \n boxes = []\n classes = []\n \n if lines[0] == '':\n return img_path, classes, boxes\n else:\n for i in range(0, int(len(lines))):\n objbud=lines[i].split(' ')\n class_ = int(objbud[0])\n \n x1 = float(objbud[1])\n y1 = float(objbud[2])\n w1 = float(objbud[3])\n h1 = float(objbud[4])\n \n xmin = int((x1*w) - (w1*w)/2.0)\n ymin = int((y1*h) - (h1*h)/2.0)\n xmax = int((x1*w) + (w1*w)/2.0)\n ymax = int((y1*h) + (h1*h)/2.0)\n \n boxes.append([xmin ,ymin ,xmax ,ymax])\n classes.append(class_)\n \n return img_path, classes, boxes\n\n\n# a function for creating file paths list \ndef create_paths_list(path):\n full_path = []\n images = sorted(os.listdir(path))\n \n for i in images:\n full_path.append(os.path.join(path, i))\n \n return full_path\n\n\nclass_ids = ['Acne']\nclass_mapping = {0: 'Acne'}","metadata":{"papermill":{"duration":0.019684,"end_time":"2024-02-03T23:31:46.311103","exception":false,"start_time":"2024-02-03T23:31:46.291419","status":"completed"},"tags":[],"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# a function for creating a dict format of files\ndef creating_files(img_files_paths, annot_files_paths):\n \n img_files = create_paths_list(img_files_paths)\n annot_files = create_paths_list(annot_files_paths)\n \n image_paths = []\n bbox = []\n classes = []\n \n for i in range(0,len(img_files)):\n image_path_, classes_, bbox_ = parse_txt_annot(img_files[i], annot_files[i])\n image_paths.append(image_path_)\n bbox.append(bbox_)\n classes.append(classes_)\n \n image_paths = tf.ragged.constant(image_paths)\n bbox = tf.ragged.constant(bbox)\n classes = tf.ragged.constant(classes)\n \n return image_paths, classes, bbox","metadata":{"papermill":{"duration":0.015092,"end_time":"2024-02-03T23:31:46.332136","exception":false,"start_time":"2024-02-03T23:31:46.317044","status":"completed"},"tags":[],"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# applying functions\ntrain_img_paths, train_classes, train_bboxes = creating_files('/kaggle/input/acne-dataset-in-yolov8-format/data-2/train/images', \n '/kaggle/input/acne-dataset-in-yolov8-format/data-2/train/labels')\n\nvalid_img_paths, valid_classes, valid_bboxes = creating_files('/kaggle/input/acne-dataset-in-yolov8-format/data-2/valid/images',\n '/kaggle/input/acne-dataset-in-yolov8-format/data-2/valid/labels')\n\ntest_img_paths, test_classes, test_bboxes = creating_files('/kaggle/input/acne-dataset-in-yolov8-format/data-2/test/images',\n '/kaggle/input/acne-dataset-in-yolov8-format/data-2/test/labels')","metadata":{"papermill":{"duration":13.457229,"end_time":"2024-02-03T23:31:59.795419","exception":false,"start_time":"2024-02-03T23:31:46.33819","status":"completed"},"tags":[],"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Creating Datasets","metadata":{"papermill":{"duration":0.005826,"end_time":"2024-02-03T23:31:59.80766","exception":false,"start_time":"2024-02-03T23:31:59.801834","status":"completed"},"tags":[]}},{"cell_type":"code","source":"# reading and resizing images\ndef img_preprocessing(img_path):\n img = tf.io.read_file(img_path)\n img = tf.image.decode_jpeg(img, channels = 3)\n img = tf.cast(img, tf.float32) \n \n return img\n\n\nresizing = keras_cv.layers.JitteredResize(\n target_size=(640, 640),\n scale_factor=(0.8, 1.25),\n bounding_box_format=\"xyxy\")\n\n# loading dataset\ndef load_ds(img_paths, classes, bbox):\n img = img_preprocessing(img_paths)\n\n bounding_boxes = {\n \"classes\": tf.cast(classes, dtype=tf.float32),\n \"boxes\": bbox }\n \n return {\"images\": img, \"bounding_boxes\": bounding_boxes}\n\ndef dict_to_tuple(inputs):\n return inputs[\"images\"], inputs[\"bounding_boxes\"]","metadata":{"papermill":{"duration":0.037818,"end_time":"2024-02-03T23:31:59.851375","exception":false,"start_time":"2024-02-03T23:31:59.813557","status":"completed"},"tags":[],"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Creating dataset loaders and tf.datasets\ntrain_loader = tf.data.Dataset.from_tensor_slices((train_img_paths, train_classes, train_bboxes))\ntrain_dataset = (train_loader\n .map(load_ds, num_parallel_calls = AUTO)\n .shuffle(BATCH_SIZE*10)\n .ragged_batch(BATCH_SIZE, drop_remainder = True)\n .map(resizing, num_parallel_calls = AUTO)\n .map(dict_to_tuple, num_parallel_calls = AUTO)\n .prefetch(AUTO))\n\n\nvalid_loader = tf.data.Dataset.from_tensor_slices((valid_img_paths, valid_classes, valid_bboxes))\nvalid_dataset = (valid_loader\n .map(load_ds, num_parallel_calls = AUTO)\n .ragged_batch(BATCH_SIZE, drop_remainder = True)\n .map(resizing, num_parallel_calls = AUTO)\n .map(dict_to_tuple, num_parallel_calls = AUTO)\n .prefetch(AUTO))\n\n\ntest_loader = tf.data.Dataset.from_tensor_slices((test_img_paths, test_classes, test_bboxes))\ntest_dataset = (test_loader\n .map(load_ds, num_parallel_calls = AUTO)\n .ragged_batch(BATCH_SIZE, drop_remainder = True)\n .map(resizing, num_parallel_calls = AUTO)\n .map(dict_to_tuple, num_parallel_calls = AUTO)\n .prefetch(AUTO))","metadata":{"papermill":{"duration":8.623777,"end_time":"2024-02-03T23:32:08.481113","exception":false,"start_time":"2024-02-03T23:31:59.857336","status":"completed"},"tags":[],"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# a function to visualize samples from a dataset\n\ndef visualize_dataset(inputs, value_range, rows, cols, bounding_box_format):\n inputs = next(iter(inputs.take(1)))\n images, bounding_boxes = inputs[0], inputs[1]\n \n keras_cv.visualization.plot_bounding_box_gallery(\n images,\n value_range=value_range,\n rows=rows,\n cols=cols,\n y_true=bounding_boxes,\n scale = 6,\n font_scale = 0.8,\n line_thickness=2,\n dpi = 100,\n bounding_box_format=bounding_box_format,\n class_mapping=class_mapping,\n true_color = (192, 57, 43))","metadata":{"papermill":{"duration":0.014839,"end_time":"2024-02-03T23:32:08.503229","exception":false,"start_time":"2024-02-03T23:32:08.48839","status":"completed"},"tags":[],"trusted":true},"execution_count":null,"outputs":[]}]} \ No newline at end of file diff --git a/ReadME.md b/ReadME.md new file mode 100644 index 00000000..b1d1ec50 --- /dev/null +++ b/ReadME.md @@ -0,0 +1,62 @@ +<<<<<<< HEAD +### Acne Prediction using Deep Learning + +#### Overview: +This project aims to predict the likelihood of developing acne using deep learning techniques. Acne is a common skin condition that affects millions of people worldwide, and early detection can aid in preventive measures and timely treatment. + +#### Dataset: +The dataset used for training and testing consists of images of facial skin affected by acne, labeled with corresponding acne severity levels. Additionally, demographic and lifestyle factors such as age, gender, diet, skincare routine, and environmental conditions may also be incorporated into the dataset for more accurate predictions. + +- **Training Dataset:** This subset of the dataset is used to train the deep learning model. It comprises a large number of labeled images along with associated non-image features. + +- **Validation Dataset:** This dataset is used to fine-tune the model hyperparameters and evaluate its performance during training. It helps prevent overfitting by providing an independent set of data for validation. + +- **Testing Dataset:** After training the model, it is evaluated on this dataset to assess its generalization performance. The testing dataset is not used during training or validation to ensure unbiased evaluation. + +#### Model Architecture: +The deep learning model employed for acne prediction utilizes convolutional neural networks (CNNs) for image processing and feature extraction. The model may also include recurrent neural networks (RNNs) or fully connected layers to incorporate non-image features from the dataset. Transfer learning techniques can be applied using pre-trained models such as VGG, ResNet, or EfficientNet to enhance performance, especially with limited data availability. + +#### Training Process: +The dataset is split into training, validation, and testing sets to train and evaluate the model. Data augmentation techniques such as rotation, scaling, and flipping may be applied to increase the diversity of training samples and improve the model's generalization. Hyperparameter tuning, including learning rate, batch size, and model architecture, is conducted to optimize performance. + +#### Evaluation Metrics: +The performance of the model is evaluated using metrics such as accuracy, precision, recall, and F1 score. Additionally, receiver operating characteristic (ROC) curves and area under the curve (AUC) values may be utilized to assess the model's discriminatory power and performance across different threshold levels. + +#### Deployment: +Once trained and evaluated, the model can be deployed as a web application, mobile application, or integrated into existing healthcare systems. Users can input relevant information such as age, gender, and upload facial images for real-time acne prediction. The application provides predictions along with confidence scores and recommendations for preventive measures and skincare routines based on predicted acne severity levels. + +#### Future Enhancements: +- Incorporating additional data sources such as genetic predisposition, hormonal factors, and medical history to improve prediction accuracy. +- Developing a user-friendly interface with interactive features for personalized skincare recommendations and tracking acne progression over time. +- Integrating natural language processing (NLP) techniques for analyzing skincare product reviews and recommendations for acne-prone individuals. +- Collaborating with dermatologists and skincare experts to validate model predictions and ensure clinical relevance and accuracy. + + + +#### Acknowledgments: +- Special thanks to [Kaggle](https://www.kaggle.com/) for providing the acne dataset. +- We acknowledge the contributions of the open-source deep learning community and pre-trained model developers. + +#### License: +This project is licensed under the [License Name]. See the LICENSE file for details. +======= +# Health Learning: ML and Deep Learning for Healthcare 🩺🧠 + +Health Learning is an open-source project aimed at leveraging machine learning (ML) and deep learning techniques to address various healthcare challenges. By harnessing the power of data-driven approaches, our goal is to develop predictive models, diagnostic tools, and decision support systems to improve patient outcomes, optimize healthcare delivery, and advance medical research. + +## Motivation πŸš€ + +The field of healthcare is ripe for innovation, with vast amounts of data available from diverse sources such as electronic health records, medical imaging, wearable devices, and genetic sequencing. Health Learning seeks to harness this wealth of data to tackle a wide range of healthcare issues, including disease prediction, diagnosis, treatment optimization, and personalized medicine. By democratizing access to healthcare data and cutting-edge machine learning algorithms, we aim to empower researchers, clinicians, and healthcare professionals to make data-driven decisions and drive innovation in healthcare. + +## Datasets πŸ“Š + +Health Learning provides access to a curated collection of healthcare datasets sourced from various sources, including public repositories like Kaggle. These datasets cover a broad spectrum of health-related topics, including maternal health, diabetes classification, cardiovascular disease risk factors, stroke prediction, cancer imaging, and more. Researchers and developers can explore these datasets to develop and validate machine learning models for a wide range of healthcare applications. Individual projects have their datasets mentioned in respective README.md files. + +## Contributing 🀝 + +Health Learning welcomes contributions from researchers, developers, healthcare professionals, and enthusiasts passionate about using machine learning and deep learning for healthcare. Whether you're interested in developing new models, improving existing algorithms, or curating datasets, there are plenty of opportunities to get involved. Check out our [Contribution Guidelines](CONTRIBUTING.md) to learn how you can contribute to the project. + +--- + +**Note**: Health Learning is a community-driven initiative and is not affiliated with any specific healthcare organization or institution. We strive to promote collaboration, transparency, and open exchange of knowledge for the betterment of healthcare worldwide. Join us in our mission to revolutionize healthcare through machine learning and deep learning! πŸŒπŸ’‘ +>>>>>>> origin/main diff --git a/data.yaml b/data.yaml new file mode 100644 index 00000000..d3e02319 --- /dev/null +++ b/data.yaml @@ -0,0 +1,13 @@ +train: ../train/images +val: ../valid/images +test: ../test/images + +nc: 1 +names: ['Acne'] + +roboflow: + workspace: osman-kagan-kurnaz + project: skin-detection-uvj1f + version: 8 + license: CC BY 4.0 + url: https://universe.roboflow.com/osman-kagan-kurnaz/skin-detection-uvj1f/dataset/8 \ No newline at end of file