Image Processing, Machine Learning

OpenCV + Posenet + Mobilenet – A Simple Implementation

Openpose is a good model for pose estimation but, it's comparatively slow and more taxing on the system. Well, there is a faster, smaller(but more error prone) human pose estimation based on mobilenet which uses the first 19 elements as against the usual full set. The model is some 7.5 MB as against the standard… Continue reading OpenCV + Posenet + Mobilenet – A Simple Implementation

Image Processing

Panorama Stitching with OpenCV – Stitcher

This is a simpler alternate to the long drawn panorama stitching process starting with keypoint detection and ending with perspective wrap. Here, you just initiate the object and pass the images as an array. Read the images left=cv2.imread('foto1A.png') right=cv2.imread('foto1B.png') Pass the images into an array. Theoretically, you can use any number of images. images=[] images.append(left)… Continue reading Panorama Stitching with OpenCV – Stitcher

Image Processing

Panorama Stitching with OpenCV

OpenCV3 was launched with a bombshell - Keypoint Detection Algorithms SIFT and SURF disappeared. Well, not exactly disappeared, they were moved to opencv-contrib library because those holding the patent for them didn't want any commercial usage for them. If you try to invoke SIFT or SURF(or BRIEF) it directly, you will get the error (-213:The… Continue reading Panorama Stitching with OpenCV