Computer vision, from pixels uphttps://blog.shyandram.dev/enhttps://blog.shyandram.dev/en/blog/welcome/https://blog.shyandram.dev/en/blog/welcome/What this site covers, how the tutorials are organized, and how to run the code.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/foundations/images-as-arrays/https://blog.shyandram.dev/en/learn/foundations/images-as-arrays/What a digital image really is, and how NumPy and OpenCV store it.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-01-introduction/https://blog.shyandram.dev/en/learn/image-processing/dip-01-introduction/What digital image processing is, where it came from, which parts of the electromagnetic spectrum it works with, and the roadmap of fundamental steps this series follows.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-02-fundamentals/https://blog.shyandram.dev/en/learn/image-processing/dip-02-fundamentals/How eyes and cameras turn light into numbers, how sampling and quantization shape a digital image, and the basic math every later chapter builds on.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-03-intensity-spatial-filtering/https://blog.shyandram.dev/en/learn/image-processing/dip-03-intensity-spatial-filtering/Remap pixel values with curves and histograms, then smooth and sharpen images with small kernels: the everyday toolbox of spatial-domain enhancement.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/convolution/https://blog.shyandram.dev/en/learn/image-processing/convolution/Blur, sharpen and find edges by sliding a 3×3 kernel across an image.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-04-frequency-domain/https://blog.shyandram.dev/en/learn/image-processing/dip-04-frequency-domain/How the Fourier transform turns an image into a recipe of waves, and how editing that recipe smooths, sharpens and removes periodic noise.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-05-restoration-reconstruction/https://blog.shyandram.dev/en/learn/image-processing/dip-05-restoration-reconstruction/Model how an image was damaged by blur and noise, then undo it with order-statistic, notch, inverse, Wiener and constrained least squares filters, and rebuild CT slices from projections.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-06-color/https://blog.shyandram.dev/en/learn/image-processing/dip-06-color/How color is measured and encoded, how to convert between RGB, HSI and L*a*b*, and how to transform, filter and segment color images.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-07-wavelets-transforms/https://blog.shyandram.dev/en/learn/image-processing/dip-07-wavelets-transforms/Treat every image transform as a change of basis, then build up from the DCT and Walsh–Hadamard transforms to Haar, multiresolution analysis and the fast wavelet transform.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-08-compression-watermarking/https://blog.shyandram.dev/en/learn/image-processing/dip-08-compression-watermarking/How images are squeezed into fewer bits, from entropy and Huffman codes to JPEG, JPEG 2000 and learned codecs, and how hidden marks are written into them.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-09-morphology/https://blog.shyandram.dev/en/learn/image-processing/dip-09-morphology/Probe shapes with small structuring elements: erosion, dilation, opening, closing, hit-or-miss, skeletons, reconstruction and grayscale morphology, with scikit-image code.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-10-segmentation/https://blog.shyandram.dev/en/learn/image-processing/dip-10-segmentation/Split an image into meaningful parts: find edges with gradients, LoG and Canny, link them with the Hough transform, pick thresholds with Otsu, and grow, cluster, cut and flood regions.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-11-active-contours/https://blog.shyandram.dev/en/learn/image-processing/dip-11-active-contours/Segment objects with curves that move: parametric snakes, gradient vector flow, balloon forces, and level-set methods such as geodesic active contours and Chan–Vese.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-12-feature-extraction/https://blog.shyandram.dev/en/learn/image-processing/dip-12-feature-extraction/Turn regions, boundaries and whole images into compact numbers that stay the same when the object moves, turns, grows or changes brightness, from chain codes and GLCM texture to Harris, MSER and SIFT.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/image-processing/dip-13-pattern-classification/https://blog.shyandram.dev/en/learn/image-processing/dip-13-pattern-classification/From minimum-distance and Bayes classifiers to perceptrons, backpropagation and convolutional neural networks: how a computer decides what an image or region shows.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/ml/ml-01-introduction/https://blog.shyandram.dev/en/learn/ml/ml-01-introduction/What it means for a computer to learn from data: the learning paradigms, linear and logistic regression, cost functions, gradient descent, overfitting and evaluation, built from scratch in NumPy.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/ml/ml-02-neural-networks/https://blog.shyandram.dev/en/learn/ml/ml-02-neural-networks/From a single perceptron to deep networks: activations, normalization, output layers and losses, backpropagation, optimizers, and the habits that make a network train and generalize.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/deep-learning/dl-01-deep-learning-for-image-processing/https://blog.shyandram.dev/en/learn/deep-learning/dl-01-deep-learning-for-image-processing/How deep learning reorganizes image processing, image analysis and computer vision: from the classical pattern-recognition pipeline to CNNs, Vision Transformers, self-supervised backbones and promptable foundation models.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/deep-learning/dl-02-image-classification/https://blog.shyandram.dev/en/learn/deep-learning/dl-02-image-classification/How deep networks decide what an image shows: losses and metrics, datasets, the CNN and transformer architectures and why they were designed that way, modern training recipes, pre-training, robustness, calibration and interpretability.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/deep-learning/dl-03-object-detection/https://blog.shyandram.dev/en/learn/deep-learning/dl-03-object-detection/How deep networks find and label every object in an image: box regression, IoU, anchors, NMS and feature pyramids; two-stage, one-stage, anchor-free and transformer detectors; open-vocabulary models; and how detection is measured.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/deep-learning/dl-04-image-restoration/https://blog.shyandram.dev/en/learn/deep-learning/dl-04-image-restoration/One degradation model, many tasks: how deep networks undo blur, noise, low resolution, haze, rain, snow and darkness, from SRCNN and DnCNN to Restormer, all-in-one models and diffusion priors.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/deep-learning/dl-05-low-light-enhancement/https://blog.shyandram.dev/en/learn/deep-learning/dl-05-low-light-enhancement/From histogram equalization and Retinex to curve estimation, Retinex transformers and diffusion models: how to brighten dark photos without amplifying noise, shifting colors or fooling the metrics.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/learn/deep-learning/dl-06-object-tracking/https://blog.shyandram.dev/en/learn/deep-learning/dl-06-object-tracking/How computer vision follows objects through video: re-identification, single-object trackers from correlation filters to transformers and SAM 2, and multi-object tracking with Kalman filters, Hungarian matching, end-to-end models and the MOTA/IDF1/HOTA metrics.Sun, 04 Oct 2026 00:00:00 GMThttps://blog.shyandram.dev/en/blog/introduction-to-machine-learning/https://blog.shyandram.dev/en/blog/introduction-to-machine-learning/Notes on the basics of machine learning: supervised and unsupervised learning, regression, classification and gradient descent.Thu, 01 Feb 2024 00:00:00 GMT