Data Science Marathon: 120 Projects To Build Your Portfolio

Explore a bankcard tampering detector built with computer vision to verify pan cards and detect tampered IDs by uploading an image and comparing it to the original using similarity scores.

Learn to detect pan card tampering in Google Colab by comparing an original image with a user image using structural similarity.

Develop a pan card detector with OpenCV by resizing images to 250 by 160, converting formats, and comparing original and tampered images using grayscale, ssim, thresholding, and contours.

Build a Flask app around your model by configuring environments, handling uploads, and wiring app.py, __init__.py, and views.py to render templates and run inference.

Create a web-based image classification app by building flask routes for get and post, handling image upload, resizing, grayscale conversion, and contour-based bounding boxes using opencv.

Deploy a Python app to Heroku by uploading code to GitHub, including app.py and requirements.txt, linking Heroku to the repo, and deploying from main to get the app URL.

Test and deploy a computer vision pan card detector, verify tampering results via a front-end file upload, and publish the app using GitHub and Heroku.

Learn to build and train a convolutional neural network for dog breed prediction using Keras and TensorFlow, including data loading from Kaggle, preprocessing, model building, evaluation, and predictions.

build a dog breed classification model with Keras and TensorFlow in Google Colab, using Kaggle data; create a CNN with Keras layers and train with train_test_split and labelBinarize, using Adam.

Load libraries and view the labels data, selecting three breeds. One-hot encode targets, convert images to a normalized numpy array by dividing by 255, and note shapes.

Build a CNN with conv2d and max pooling, flatten, and dense layers for a three-class image classification, using softmax output, categorical cross entropy, and Adam optimizer, with hyperparameter tuning.

Plot training and validation accuracy per epoch with matplotlib, examine history, and show how hyperparameter tuning improves accuracy to 83.33% on the test set.

Build a Streamlit app to deploy a dog breed classifier by saving the model as dog breed dot h5 and loading it with Keras, featuring interactive image upload and prediction.

Set up a data science app workflow by creating the requirements file, installing Streamlit, and running the app in an Anaconda prompt to view dog breed predictions locally.

Learn to add image and text watermarks with OpenCV in Google Colab by loading images, resizing, converting to RGB, and merging a logo and text to protect copyrights.

Learn to add a logo and text watermark to images in Google Colab using OpenCV to provide copyright protection, with NumPy and PIL for fetching, resizing, and displaying.

Apply image and text watermarking using OpenCV and NumPy: convert to rgb, compute image center, define region of interest, and overlay a logo or text watermark.

Create a flask app for the image watermarking project, configuring app settings, uploading images, and applying logo or text watermarks with cv2, PIL, numpy, and deploying to Heroku.

Deploy a Python watermarking app to Heroku from GitHub, enabling automatic deploys and installing requirements. Test the logo and text watermarking via the deployed app URL.

Build a multi-class traffic sign classifier with a CNN using TensorFlow and Keras. Learn data preparation, Kaggle data loading in Colab, model training, evaluation, and prediction visualizations for self-driving cars.

Connect your Kaggle account to Google Colab, download and unzip the German traffic sign dataset, and import libraries for building a convolutional neural network with TensorFlow and Keras.

Visualize traffic sign images, resize to 50 by 50 rgb, normalize by 255, and one-hot encode labels, then split data into 80/20 train/validation for a 43-class dataset of 39,209 images.

Design and test a cnn architecture with conv 2d, 64 filters, 3x3 kernels, maxpool, dropout, and a 43-class dense output using softmax and sparse categorical cross-entropy.

Learn to preprocess traffic sign data by scaling images, normalizing with 0–1, and training a conv2d model that achieves 95% accuracy on 43 classes for self-driving cars.

Extract text from images with Tesseract and Pytesseract, then enhance results using OpenCV techniques like blur and threshold to improve OCR accuracy.

Learn to set up Google Colab for text extraction, install Tesseract, Pytesseract, and OpenCV, and import libraries like PIL, NumPy, and re to download, preprocess, and extract text from images.

Extract text from images with Pytesseract using English, cleanse symbols, then preprocess via OpenCV: load image, convert to grayscale, denoise with blur, and apply thresholding.

Apply erosion and morphology to refine foreground, detect edges with canny, and correct skew; then use template matching and tesseract to outline words for a Flask app.

Build a flask optical character recognition app, configure settings, handle image uploads, convert images to array and grayscale, extract text with tesseract, and display results via the front end.

Install the required libraries with pip install -r requirements.txt, then start the app using python app.py. Upload an image with text to extract text via optical character recognition.

Explore building a convolutional neural network for plant disease detection using TensorFlow and Keras in Colab, including data visualization, preprocessing, one-hot encoding, model training, and predictions for agricultural impact.

Mount Google Drive in Google Colab, import essential libraries, and build a TensorFlow Keras CNN to classify plant health, training and testing with NumPy, pandas, Matplotlib, and OpenCV.

Visualize plant disease images, convert them to numpy arrays, and preprocess the data with train test split, normalization, and one hot encoding for three plant diseases.

Build a CNN for plant disease classification with conv2d, max pooling, flatten, dense layers, and softmax for three classes; train with categorical cross entropy and Adam, achieving 99.44% test accuracy.

Create a Streamlit app to deploy a plant disease classifier; load a Keras model, upload leaf images processed with OpenCV, and predict diseases like tomato bacterial spot.

Detect and count vehicles in images or video using OpenCV and a Haar cascade classifier loaded from an XML file, with grayscale conversion and contours for labeling and location.

Learn to detect and count vehicles in images and videos using Google Colab, OpenCV, and Haar cascade classifiers (cars.xml, bus_front.xml), with code for loading images and processing.

Convert images to grayscale, apply Gaussian blur, dilate, and perform morphological operations with kernels in OpenCV; detect cars and buses with Haar cascades and count objects in video.

Build a Flask app for vehicle detection and counting using OpenCV cascade classifier and image uploads, with deployment on Heroku via GitHub.

Learn to build a face swapping pipeline with OpenCV and Dlib. Use facial landmarks, triangulation, masks, and seamless clone for realistic color matching.

Utilize Google Colab to run a face swapping project with OpenCV and Dlib, loading a pre-trained shape predictor 68 face landmarks model, and processing images from URLs.

Convert images to numpy arrays and grayscale, detect 68 facial landmarks with dlib, triangulate and warp triangles, then use seamlessClone to finalize the swapped face.

Build a Flask faceswap application from config to deployment, detailing app structure, views, and pre-trained models with dlib. Design the front end with templates and the requirements.txt setup.

Build a CNN with Keras to predict bird species from images in Google Colab, covering data preparation, model training, testing, and result visualization.

build and train a convolutional neural network in google colab to classify bird species using keras, tuning activation functions, learning rate, batch size, and epochs on data loaded from drive.

Assess data balance across six bird species with 811 labels, using 224 by 224 RGB images; apply 80/20 train-test split, normalize by 255, reshape, one-hot encode, and create validation split.

Build a cnn model using a sequential architecture with conv 2D, max pooling, flatten, and dense layers, trained with adam, dropout, and categorical cross entropy to classify bird species.

Explore how to create a Flask app for a six-class bird species classification model, configure settings, load a pre-trained model, handle uploads, and display predictions.