Here I build Deep Learning, Computer Vision, NLP, General Machine Learning applications/tutorials while maintaining code readability. If you're looking for learning ML resources, this is what you are looking for.

• Image filtering: To blur, add/remove nosie, sharpen image.
• Edge detection: Gradient Operators, Laplacian of Gaussian, Canny Edge Detection.
• Corner Detection: Harris, Shi-Tomasi, FAST Corner Detection.
• Keypoints detection and matching: Using ORB to detect keypoints and descriptors, then match with other images.
• Image Pyramid And Reconstructing Image: Gaussian, Laplacian Pyramid and Image reconstruction.
• Optical flow demo notebook[Extra]: Little demo.
• Optical flow: Find flow of particular object.
• Dense flow: Find global motion, process is to track all pixels to estimate motion.
• Bag of visual words: Make BOVW features to train an image classifier on.

• Face & Facial landmarks Detection: Using Haarcascades, Face Detection library, dlib.
• Face recognition using dimension reduction: Make features using PCA/LDA which are dimension reduction algorithms and then use a classifier on top to classify the face.
• Face_recognition using deep learning: A more modern and prefered way using deep learning library for faces called Face Detection library.
• Extract faces[Extra]: Extract faces from a folder just by running this .py file into the folder containing images using Haarcascades.

• Environment Sound Classification: Using spectorgrams as features to train image classifier.
• Image Classification:
  1. Image Classfication using BOVW: Use Bag Of Visual Words(BOVW) features to classify images using SVM.
  2. Image Classification using HOG: Use Histogram Of Gradients Features to classify images using SVM.
  3. Using Deep learning + Clean pipeline: Using resnet18 to classify flowers, also a clean pytorch training pipeline.
• Object Detection:
  1. Object Detection Using HOG(part 1): Train car/non-car using HOG featues and SVM classifier.
  2. Object Detection Using HOG(part 2): Create object detector with sliding window approach.
  3. Object Detection Using Faster R-CNN(part 1 optional): Train a resnet18 car/non-car classifier.
  4. Object Detection Using Faster R-CNN(part 2): Train Faster-RCNN object detector with resnet car classifier as backbone.
• Image Segmentation:
  1. Semantic Segmentation Using U-Net: Semantic Segmentation using U-Nets on Oxford-Pets dataset.
• Auto-Encoders:
  1. Image denoising using AutoEncoders: Image reconstruction and image denoising using autoencoders.
  2. Image coloring using AutoEncoders(RGB image): Image colorization using cnn autoencoders.
  3. Image coloring using AutoEncoders(LAB image): Image colorization using cnn autoencoders.
• Image Translation:
  1. Unpaired Image-Image Translation: Unpaired Image-Image translation using CycleGans.

• Knowledge Distillation pytorch: Train Teacher model, then train Student model with same or smaller model size. Also Improves generalization.
• Stochastic Weight Averaging pytorch: Improve generalization by Weight Averaging after training, by training for ~25% more with SWA.
• Gradient accumulation pytorch: Update weights with bigger batch size, by updating weights after some number of iteration.
• Automatic Precision Training pytorch: Fit a bigger batch, improve training speed using Automatic Precision(Mixed Precision Training), basically doing multiplications with fp16 and addition with fp32.

• Petals to the Metal: FLower Classification on TPU: Image classification using effnets on TPU with Tensorflow.
• Wheat Detection: Object detection using efficient-det and faster-rcnn with Pytorch.