/Robomaster_NTU_2018

Robomaster vision

Primary LanguagePython

Robomasters Software Team

This repo is created for Robomasters 2018. Software Team of MECATron, Nanyang Technological University.

Tutorials

As we are going to implement computer vision and deep learning techniques in system, we firstly conduct several tutorials.

The marked items are finished topics and unmarked ones are still in progress or in plan.

The tutorials on deep learning basics:

  • Tensorflow and sul

  • Regression (Linear and Logistic)

  • MLP (Multi-layer perception)

  • Convolutional Network

The tutorials on computer vision:

  • Classification with CNN

  • Object detection with CNN

  • Object Detection with FCN (Fully Convolutional Network)

  • Optimization on networks

The tutorials on reinforcement learning:

  • Value and policy

  • Q learning and Deep Q learning

  • Tricks on game state search

The tutorials on data analysis:

  • Data properties and concepts on feature domain.

  • Data collection and augmentation.

  • Pre and post process of collected data.

ROS tutorial:

  • ROS tutorial

We dont need to use ROS for this simple project. We do it on our own.

Practice

  • autoboot_test: Auto boot script

  • model_zoo: Models of different training set. Old files just for archive

  • New_protocol: Deploy codes for different kinds of robots.

  • record_video: Recording script.

  • training_code: Code for training.

  • icra_robot_system: Codes for ICRA 2018.

Lib

You must know what this means.

Label

Label tool, with instructions.

Extract frames

Script to extract frames from video.