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