DEV Community: Ibra

License plate cover on car images

Ibra — Thu, 08 Sep 2022 20:11:00 +0000

Introduction

Goal of this project was to develop and train a Deep learning model capable of detecting and covering license plates of vehicles.

Technical

This project was implemented with Python using the Keras framework on top of Tensorflow.
A number of python libraries like OpenCV, Numpy, Pandas, etc… were used.

Examples

Deployment

A lot of work done in order the reduce the overall size of the code and model weights to allow a deployment in Serverless environments like AWS Lambda and Google Cloud Functions.

Finally i was able to deploy on Google Cloud Function with an API in front.

The serverless approach is very optimal as the scalability is handled for you and serverless is almost always cheap.
In my case, i had a deep learning model that was almost totally free (1 million operations per month for free).

Source:
https://www.ibragim.fr/portfolio/cache-plaque-deep-learning/

Background removal on car images - Machine learning project

Ibra — Thu, 08 Sep 2022 06:35:40 +0000

Introduction

The goal of this project was to have an algorithm capable of removing the background of a vehicle image automatically with machine learning.

Technical

This was done with Python using the framework Keras, Tensorflow and a number of known python libraries like Numpy, Pandas, OpenCV, etc…

The algorithm uses a lot of CPU power to predict and remove the background and by consequence, this raises a scalability issue, it can get very costly to deploy and scale something like this.
So to get something high-performance and scalable, i did a lot of work optimizing the code, packaging, removing useless code all to reduce the size of the end result.

After weeks of hard work, i was able to deploy it in a Serverless environment on Google cloud with Cloud Functions.
So now, i had a ML model that is fast, scalable and almost totally free (1mil operations per month for free).

Source:
https://www.ibragim.fr/portfolio/detourage-automatique-voitures-ia/

Batch processing with Symfony and Docker

Ibra — Mon, 05 Sep 2022 16:07:01 +0000

Context

We had a platform running with Symfony where we needed to execute a lot of heavy background jobs daily.

Initially we used a bundle which handles our jobs and their execution.
With the growing user base, it had started to let us down.
We were getting more and more jobs that are stuck, hung, etc…

Solution

To solve this, i developed a new bundle that offers a batch processing system with the Docker engine.

All jobs are executed separately in docker containers.

This way is more viable as Docker handles the execution of the started containers.

Neat thing with this bundle is, you can pretty much execute anything in those containers.
For example, say you need a downloader in your symfony application, rather than implementing it in PHP, you could leverage JS Async and make a quick node.js script with something like bluebird and package it into a Docker image.

Monitoring UI

The bundle also offers a Monitoring user interface which you may choose to activate.
With this interface, you can:
— get an overview of your jobs
— inspect jobs
— stop jobs
— get usage info for each running job

Job details page

Dashboard page

Check out the bundle, see if it's what you needed.

Github: https://github.com/ibra-akv/docker-jobs-bundle