Object Detection for Additive Manufactured Parts

The rest of this repository contains the code base used on the article(s) to be published.

Setup

Installing CUDA 10.1

You will need CUDA 10.1 in order to run mxnet 1.51, on Ubuntu:

On Ubuntu 20.04

sudo apt update
sudo apt install nvidia-cuda-toolkit

On Ubuntu 18.04

Add Nvidia repository:

sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub && echo "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64 /" | sudo tee /etc/apt/sources.list.d/cuda.list

Install CUDA 10.1

sudo apt-get update && sudo apt-get -o Dpkg::Options::="--force-overwrite" install cuda-10-1 cuda-drivers

Setting up the environment:

Create a environment (I use conda):

conda create -n traindet python=3.7.6
conda activate traindet

Install packages:

git clone https://github.com/czrcbl/train_detection
cd train_detection
pip install -r requirements.txt

In order to run some scripts, you need to install the following packages:

git clone https://github.com/czrcbl/bboxes
cd bboxes
pip install -e .

git clone https://github.com/czrcbl/detection
cd detection
pip install -e .

Rendering

In order to run the rendering code, you need the blender executable, version 2.80, in the path. Download from HERE.

Structure

• Folder traindet has the core utilities from training and evaluating the models.
• Folder rendering has all the code that is supposed to be run through blender. You must have the blender executable on system path.
• Folder scripts has the scripts for launching the networks training, starting the rendering of synthetic images and producing some visualizations.
• All the data necessary to train the models should be placed on the data folder.

Training a model

Just call the model train scrip with the required arguments, models are saved under data/chackpoints/<dataset_name>/<model_name>: Some example calls are on the command.py file.