nyakasko Goto Github PK

anomalydetection

In this assignment I have to build a reconstruction loss based anomaly detection model using tf.keras. I have to train an autoencoder to reconstruct Homer Simpson images, then use the trained model to classify Simpson and Flower images.

automatic_scripts

computervision-ransac

Fit RANSAC lines on edge detected images - Sequential RANSAC

darknet

YOLOv4v / Scaled-YOLOv4 - Neural Networks for Object Detection (Windows and Linux version of Darknet )

homography_estimation

Feature matching (SIFT) between two images and then applying normalized linear homography estimation, robustified by standard RANSAC

icp_tricp

Implementation of Iterative Closest Point and Trimmed Iterative Closest Point algorithms.

imageandvideoanalysis

Prewitt edge detector: gradient filter és nonmaxima-suppression (NMS), Thresholding algorithm by Otsu, Detection of circular object by edge detection and Hough transform for circles, Motion tracking of feature points and dense optical flow

imagefiltersandupsampling

Various image filters and Guided Joint bilateral filter for guided image upsampling

lo-ransac-3dplanefitting

Implementation of the Locally Optimized Random SAmple Consensus (LO-RANSAC) 3D plane fitting algorithm

mlquestions

Machine Learning and Computer Vision Engineer - Technical Interview Questions

objectandkeypointdetection

In this assignment I have to build a Mask R-CNN based keypoint detector model using Detectron2. Detectron2 was written in PyTorch and contains many state-of-the-art obejct detection models with pretrained weights.

stereo_reconstruction

Estimating the fundamental and essential matrices of input stereo images, and then reconstructing the 3d points by triangulation.

stereomatching

Disparity image creation with naive and dynamic programming approaches and 3D pointcloud generation from the disparity image

zed-examples

ZED SDK Example projects

zed-opencv

ZED SDK interface sample for OpenCV

zed-pcl

ZED SDK interface sample for PCL, the Point Cloud Library

zed-pytorch

3D Object detection using the ZED and Pytorch

zed-tensorflow

3D Object detection using the ZED and Tensorflow

zed_2_change_detection

This code is an initial version of the change detection algorithm developed using the ZED 2 stereo camera. The program was developed by Levente Göncz at SZTAKI (Institute for Computer Science and Control) in a framework of a 4 months long project.