https://github.com/mastergalen/motion-relationships

UCL Computer Science MEng Master's Project: Using machine vision to predict motion relationships between entities in videos.
https://github.com/mastergalen/motion-relationships

computer-vision keras machine-vision python ucl

Last synced: 16 days ago
JSON representation

UCL Computer Science MEng Master's Project: Using machine vision to predict motion relationships between entities in videos.

• Host: GitHub
• URL: https://github.com/mastergalen/motion-relationships
• Owner: Mastergalen
• Created: 2017-12-22T18:01:07.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2018-04-30T17:07:20.000Z (almost 7 years ago)
• Last Synced: 2024-11-21T11:41:25.633Z (3 months ago)
• Topics: computer-vision, keras, machine-vision, python, ucl
• Language: Python
• Homepage:
• Size: 7.45 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Motion Relationships

This repository contains all the code for my Master's thesis at UCL: **Predicting motion relationships in video**.

The report can be viewed [here](https://cdn.galenhan.com/documents/university/thesis/Thesis-Predicting_Motion_Relationships_in_Video.pdf).

![Motion class example](figures/motion-class-examples.jpg)

## Prerequisites

* Copy `.env.example` to `.env`

* Fill in API keys

* Install Python packages `pip install -r requirements.txt`

## Annotation Interface

The code for the annotation interface used for collecting annotations on AMT can be found inside `annotation-ui`.

## Scripts

Executable scripts for various pre-processing tasks can be found within the `scripts/` directory.

Each script is individually documented.

## Models

To train/run the motion relationship models see `lib/model/architectures`.

## Dataset

The annotated dataset can be downloaded [here](https://cdn.galenhan.com/documents/university/thesis/motion-relationship-dataset.zip).