creative-optimisation-cv's Introduction

Welcome to Creative optimisation using computer vision algorithms 👋

Project Description 📖

An image of an Ad creative is used as input for a feature extraction pipeline. The pipeline extracts important features from this image. An Ad creative can contain the following elements:

Text
Colour
Objects
Logo
Faces
CTA button
Engagement button

Details such as the object location, emotions, button location, colour etc. are features that can all be derived from an image using the pipeline. After feature extraction, a regression model is able to determine which features lead to an increase in KPI. The KPIs used in this project are:

Engagement Rate
Click Through Rate

This project has applications in the advertising industry where Ads need to be personalised and contain the right content.

Project Workflow Diagram 📋

Project Structure 📁

├─ .DS_Store
├─ .github
│  └─ workflows
│     └─ main_model.yaml
├─ .gitignore
├─ LICENSE
├─ README.md
├─ models
│  ├─ build_model.py
│  └─ model_training.png
├─ notebooks
│  ├─ Aesthetic_Features.ipynb
│  ├─ AssetExtractionExtraction.ipynb
│  ├─ DataExtractor.ipynb
│  ├─ FeatureExtraction.ipynb
│  ├─ FeatureSelection.ipynb
│  ├─ ModelDevelopment.ipynb
│  ├─ text_detection.ipynb
│  └─ web_scrab.ipynb
├─ requirements.txt
└─ scripts
   ├─ CTA_extract.ipynb
   ├─ color_comp_texture.py
   ├─ extract_url.ipynb
   ├─ extractor_pipeline.py
   ├─ feature_extractor.py
   ├─ matching_detector.py
   ├─ multiple_face_detect.py
   ├─ object_detector.py
   ├─ select_asset_id.py
   ├─ text_detect.py
   ├─ unzip.py
   └─ url_to_list.ipynb