Devising Deep learning method for the classification of garbage.

I am main developer driving the whole AI ecosystem from development till deployment

The requirement of the project was that all the phishing data having millions of records is to be managed, easily queried with certain tags and downloaded with ease over the cloud space. I was the main developer of the project. I carried out following steps in developing the solution using ELK stack :

1- Acquire Test VM for setting up the ELK stack
2- Configuring the ELK stack to the Test VM and test upload sample data
3- Setting the pipeline for uploading the original data
4- Coding of the data uploading, downloading and updating python scripts
5- Run different queries for testing query turnaround time, query failures (if any), batch downloading speed+time and storage estimation
6- Acquiring the production grade VMs and set up the cluster
7- Replicating the system with single VM or cluster (depends upon the requirement)
8- Run queries and run tests for the total operations

This system contained 3D convolution deep learning model trained for IP cams ofUET. They had normal classes as well as suspicious ones like gun pointing andfighting. I was leading the project with two resources. We did following steps for the development of project:

1. Data Collection and Preparation
2. Data Annotation
3. Data Preprocessing and Augmentation
4. Model Defination
5. Model training
6. Model Testing
7. Model Deployment

Developing a complete machine learning solution from development to production thatcould detect and classify the garbage type into 17 distinct categories.Data gathering, preparation and analysisConduct experiments to come up with best performance model.

Processed data is to be managed in such a way that it can be queried by whole datascience team with certain tags.Data bank management operations are to be done using MySQL and python.Queries and data insights using ELK stack.Develop prediction models on curated Tabular and NLP datasets

Trained for IP cams of educational institute.Detect normal classes as well as suspicious ones like gun pointing andfighting.The model was deployed and encapsulated inside on Django backend service

Facial Analysis using conventional and advance computer vision techniques