Skip to content

qimaqi/Computer_Vision_Proj_public

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Computer vison course projects

This branch collect all the code assignment as well as the report of the ETHz course Computer vison. I get full grade for all these projects.

  • Computer vison assignments
    • exercise1: Introduction
      • Introduction to Pytorch
    • exercise2: Image Segmentation
      • Mean-Shift
      • SegNet
    • exercise3: Camera calibration and structure from motion
      • DLT Algorithm
      • Structure from motion
    • exercise4: Model fitting and multi-view vision
      • Model Fitting
      • Stereo/Multi-view vision
    • exercise5: Object Recognition
      • Bag-of-words Classifier
      • CNN-based Classifier
    • exercise6: Tracking
      • CONDENSATION: Tracker Based On Color Histograms

To protect the property of the teaching staff I only show my reports here, not the code

exercise1: Introduction

Introduction to Pytorch, A demo for getting used to pytorch modules

exercise2:Image Segmentation

Mean-Shift

I use the mean shift algorithm with gaussian kernel with 2.5 bandwidth and get following results:

drawing

drawing

SegNet

I use the SegNet(Like UNet) and get following results:

drawing

exercise3: Camera calibration and structure from motion**

DLT Algorithm

The reprojection error is 6.253e-4

drawing

Structure from motion

The reprojection error is 6.253e-4 I use the global SFM instead of the sequential SFM get better resutls.

drawing

drawing

exercise4: Model fitting and multi-view vision

Model Fitting

I implement the ransac line fitting alogrithm and is robust to the outlier of these yellow points.

drawing

Stereo/Multi-view vision

I implement the MVS Net for the dense depth estimation task and get good results.

drawing

drawing

exercise5: Object Recognition

Bag-of-words Classifier

In this task I solve the car recognition task using bag-of-words classifier and conduct the experiment about how different cluster number will influence the results.

drawing

drawing

CNN-based Classifier

In this task I solve the CIFAR-10 Dataset using simplified VGG net and achieve 82% accuracy.

drawing

drawing

exercise6: Tracking

CONDENSATION: Tracker Based On Color Histograms

In this task I use the particle tracking alogrithm based On Color Histograms. Below are the results for different tracking challenge.

drawing

drawing

drawing

Conclusion

I really learned a lot during this courses and will keep going forward to dive more in the topic of computer vision and deep learning

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published