Online Ad Creatives With Sizes And Extracted Text

9,003 samples of online programmatic ad creatives include ad sizes and extracted creative text. The dataset, created by PeterBrendan, contains various ad dimensions like (300, 250) and (728, 90). It was last updated in August 2024.

Use Cases

• Train a multi-modal model to associate creative text with specific ad sizes like (970, 250) or (300, 600).
• Analyze the distribution and frequency of ad size categories across the 9,003 creative samples.
• Build a classifier to predict ad size from extracted creative text features.
• Use the image and text pairs for retrieval tasks matching visual ad creatives to their textual content.

Strengths

• Contains 9,003 samples of ad creatives, providing a substantial collection for analysis.
• Includes multiple specific ad size dimensions, such as (300, 250) and (728, 90), as key features.
• Combines both image and text modalities (creative and extracted text) for each sample.
• Dataset was updated in August 2024, indicating recent maintenance.

Limitations

• The sample size of 9,003 may be limited for training large-scale generative models.
• Geographic and platform source bias is likely, as data appears curated from a specific larger source (Regionus tag).
• Lacks detailed column definitions and sample data, limiting immediate understanding of data structure.

Provenance

Source

Curated from a larger source of online programmatic ads by author PeterBrendan.

Collection Method

Curated from a larger collection; specific gathering method not detailed.

Freshness

Last updated on 2024-08-21.

Geography

Likely United States focus (based on 'Regionus' tag).