dgtl.vsn

An object detection model for digital user interfaces

Installation

Fresh Docker Container setup

1. docker run --gpus all -it -p 8888:8888 --name dgtlvsn tensorflow/tensorflow:latest/gpu
2. apt-get update && apt-get install git
3. apt-get install protobuf-compiler
4. apt-get install wget
5. python -m venv my-env
6. source my-env/bin/activate
7. python -m pip install --upgrade pip
8. pip install ipykernel
9. python -m ipykernel install --user --name=my-kernel
10. python -m pip install --upgrade jupyter

Existing python environment

11. Clone this repository.
12. Open 1_Setup.ipynb to install packages, creates folders, and ensure the check is satisfied.
13. Download sample dataset or place images in /Tensorflow/workspace/images/train and /Tensorflow/workspace/images/test.

Result Summary

Training and evaluation were performed within a Docker container on a Windows machine (16gb RAM, RTX 3080)

Model: SSD Mobile Net 640x640

Sites	Images	Steps	AP	AR	CL	TL
5	10	2000	0.839	0.421	0.0995	0.4056
5	10	5000	0.872	0.429	0.1002	0.2400
5	50	2000	0.871	0.489	0.0260	0.1722
5	50	5000	0.896	0.498	0.0292	0.1540
5	50	10000	0.966	0.539	0.0213	0.1149
5	100	2000	0.909	0.452	0.0639	0.2225
5	100	5000	0.927	0.464	0.0184	0.1463
5	100	10000	0.879	0.439	0.0267	0.1220