/Computer-vision

Computer vision week at Akademy.AI bootcamp.

Primary LanguageJupyter Notebook

Computer Vision Week

Week 7 of the Akademy.AI bootcamp of summer 2019 was dedicated to computer vision. Here is an overview of the week's projects and subjects.

List of completed projects in CV

1. Lane detection for self-driving cars

img1

2. Ball tracker

img2

3. Handwritten digits recognition using own pictures

img3

4. Identifier for triangles, rectangles and polygons

img4

5. Noise reducer for photos

img5

6. Text enhancement

img6

7. Live Art Filter

img7

8. Meme generator

img8

New Computer Vision skills, usages and importance

  1. OpenCV basics (loading, storing, converting channels, cropping and plotting images): important to be fluent in these tasks for all further CV work
  2. Drawing shapes (lines, rectangles, circles, ellipses, poligons, text, custom fonts, overlays): useful for many tasks, such as creating masks and drawing lines when detecting patterns in images or video
  3. Color spaces (splitting and merging color channels, HSV): useful for changing the hue/saturation/value of an image, or enhancing a particular color channel, as well as brightening and darkening images
  4. Working with video (opening up a video capture and using the webcam or a video file): essential when working with any kind of video.
  5. Affine transformation (translation, rotation, scaling, upsampling/downsampling images): important preprocessing steps
  6. Image pyramids (gaussian, laplacian): for image blending
  7. Bitwise operations (AND, OR, NOT, XOR): useful for masking
  8. Blurring and sharpening (gaussian blur, median blur, bilateralFilter...): for preprocessing to eliminate noise or enhance features
  9. Binarization (thresholding) (binary, binary inverted, adaptive thresholding, etc): very important when we want just the light or the dark areas in the image
  10. Morphological transformations (opening - erosion followed by dilation, closing - dilation, then erosion): for noise removal
  11. Edge detection and perspective: for example for aligning skewed images of documents
  12. Contours (detecting and sorting edges of objects, approximation, convex hull, fitting rectangle/ellipsis, minimum enclosing circle/triangle, contour shape matching, line detection, Hough lines, circle detection): useful for object detection
  13. Object detection (template matching, corners, (SIFT, SURF, FAST,) ORB, HAAR cascades: detect certain features in objects, such as patterns or facial features, and can match them to new images

In-class contributions

  1. Brought cake on Friday!
  2. Participated in class challenges
  3. General joy to be around 😝