iscc/iscc-book-covers

---
annotations_creators:
- machine-generated
language_creators:
- found
language:
- en
license: other
multilinguality:
- monolingual
size_categories:
- 1M<n<10M
source_datasets:
- cogsci13/Amazon-Reviews-2023-Books-Meta
task_categories:
- image-feature-extraction
- zero-shot-image-classification
tags:
- iscc
- content-identification
- similarity-search
- deduplication
- image
- iso-24138
- amazon
- books
- book-covers
pretty_name: "ISCC Codes for Amazon Book Covers"
dataset_info:
features:
- name: image_url
dtype: string
- name: iscc
dtype: string
- name: iscc_meta
dtype: string
- name: iscc_semantic
dtype: string
- name: iscc_content
dtype: string
- name: iscc_data
dtype: string
- name: iscc_instance
dtype: string
- name: source_row_id
dtype: string
- name: title
dtype: string
- name: isbn
dtype: string
- name: publisher
dtype: string
---

# ISCC Codes for Amazon Book Covers

Amazon book covers enriched with full 256-bit ISCC (International Standard Content Code) identifiers for cover image identification, similarity search, and deduplication research. The `image_url` field links to cover images on Amazon CDN for preview.

## What is ISCC?

The **International Standard Content Code** ([ISO 24138:2024](https://www.iso.org/standard/77899.html)) is a
content-derived identifier for digital media assets. Unlike traditional identifiers that are assigned
arbitrarily, ISCC codes are generated algorithmically from the content itself, enabling:

- **Content Identification**: Identify content regardless of format or location
- **Similarity Search**: Find visually or semantically similar images
- **Deduplication**: Detect exact and near-duplicate content
- **Provenance Tracking**: Link derived works to their sources

## ISCC Units

Each record contains five 256-bit ISCC-UNITs that capture different aspects of the content:

| Unit | Field | Description |
|------|-------|-------------|
| **Meta-Code** | `iscc_meta` | Similarity based on embedded metadata (filename, title) |
| **Semantic-Code** | `iscc_semantic` | AI-based visual semantic similarity (what the image depicts) |
| **Content-Code** | `iscc_content` | Perceptual image similarity (visual appearance) |
| **Data-Code** | `iscc_data` | Raw binary data similarity (file structure) |
| **Instance-Code** | `iscc_instance` | Cryptographic hash for exact matching (like SHA-256) |

The `iscc` field contains the composite ISCC-CODE combining all units.

## Dataset Structure

### Data Fields

| Field | Type | Description |
|-------|------|-------------|
| `image_url` | string | Cover image URL on Amazon CDN |
| `iscc` | string | Full composite ISCC-CODE |
| `iscc_meta` | string | 256-bit Meta-Code |
| `iscc_semantic` | string | 256-bit Semantic-Code |
| `iscc_content` | string | 256-bit Content-Code |
| `iscc_data` | string | 256-bit Data-Code |
| `iscc_instance` | string | 256-bit Instance-Code |
| `source_row_id` | string | Original row identifier (`parent_asin`) in source dataset |
| `title` | string | Book title |
| `isbn` | string | ISBN-13 (preferred) or ISBN-10 |
| `publisher` | string | Publisher name and edition info |

### Data Splits

| Split | Samples |
|-------|---------|
| train | 3,079,720 |

## Usage

### Loading the Dataset

```python
from datasets import load_dataset

ds = load_dataset("iscc/iscc-book-covers")
```

### Viewing a Sample

```python
sample = ds["train"][0]
print(f"ISCC: {sample['iscc']}")
print(f"Title: {sample['title']}")
```

### Similarity Search Example

```python
import iscc_core as ic

# Get two ISCC codes to compare
code1 = ds["train"][0]["iscc_content"]
code2 = ds["train"][1]["iscc_content"]

# Calculate hamming distance (0 = identical, 256 = maximally different)
distance = ic.iscc_distance(code1, code2)
print(f"Hamming distance: {distance}")

# Convert to similarity percentage
similarity = 1 - (distance / 256)
print(f"Similarity: {similarity:.1%}")
```

### Finding Near-Duplicates

```python
import iscc_core as ic

# Threshold for near-duplicates (adjust based on use case)
THRESHOLD = 32 # ~87.5% similarity

reference = ds["train"][0]["iscc_content"]

for i, row in enumerate(ds["train"]):
distance = ic.iscc_distance(reference, row["iscc_content"])
if distance <= THRESHOLD and i > 0:
print(f"Near-duplicate found: row {i}, distance={distance}")
```

### Semantic Similarity Search

```python
import iscc_core as ic

# Find semantically similar images (same subject/concept)
reference = ds["train"][0]["iscc_semantic"]

similar = []
for i, row in enumerate(ds["train"]):
distance = ic.iscc_distance(reference, row["iscc_semantic"])
if distance <= 64: # ~75% semantic similarity
similar.append((i, distance))

# Sort by similarity
for idx, dist in sorted(similar, key=lambda x: x[1])[:5]:
print(f"Row {idx} (distance={dist})")
```

## Source Data

This dataset was derived from [cogsci13/Amazon-Reviews-2023-Books-Meta](https://huggingface.co/datasets/cogsci13/Amazon-Reviews-2023-Books-Meta).

### Source Data
Book metadata from the Amazon Reviews 2023 dataset by McAuley Lab.
Cover images hosted by Amazon CDN. This derivative dataset contains
ISCC codes and references to the original images, not the images themselves.
**License**: Research use only. Refer to original dataset terms.

### Processing

ISCC codes were generated using:
- [iscc-sdk](https://github.com/iscc/iscc-sdk) - High-level ISCC generation
- [iscc-sci](https://github.com/iscc/iscc-sci) - Semantic image codes (experimental)

All processing was performed on original resolution images downloaded from Amazon CDN.

## Considerations

### Intended Use

- Content identification and matching research
- Image similarity search algorithm development
- Deduplication system benchmarking
- Visual-semantic retrieval experiments
- ISCC-based indexing research

### Limitations

- Semantic codes are generated using experimental AI models and may not capture all semantic nuances
- ISCC codes are sensitive to significant image modifications (heavy cropping, overlays, filters)
- Image URLs point to Amazon CDN and may become unavailable over time

### Privacy

This dataset contains ISCC codes and image URL references derived from the source dataset. Refer to the original
dataset documentation for privacy considerations.

## Citation

If you use this dataset, please cite both this dataset and the original source:

**This Dataset:**
```bibtex
@dataset{iscc_book_covers,
title = {{ISCC Codes for Amazon Book Covers}},
author = {{ISCC Foundation}},
year = {{2026}},
publisher = {{Hugging Face}},
url = {{https://huggingface.co/datasets/iscc/iscc-book-covers}}
}
```

**Amazon Reviews 2023:**
```bibtex
@article{hou2024bridging,
title = {{Bridging Language and Items for Retrieval and Recommendation}},
author = {{Hou, Yupeng and Li, Jiacheng and He, Zhankui and Yan, An and Chen, Xiusi and McAuley, Julian}},
journal = {{arXiv preprint arXiv:2403.03952}},
year = {{2024}}
}
```

**ISCC Standard:**
```bibtex
@misc{iso24138,
title = {{ISO 24138:2024 Information and documentation -- International Standard Content Code (ISCC)}},
author = {{International Organization for Standardization}},
year = {{2024}},
url = {{https://www.iso.org/standard/77899.html}}
}
```

## Additional Resources

- [ISCC Foundation](https://iscc.io/) - Standards organization
- [ISCC Documentation](https://sdk.iscc.codes/) - Technical documentation
- [ISO 24138:2024](https://www.iso.org/standard/77899.html) - Official standard
- [iscc-sdk](https://github.com/iscc/iscc-sdk) - Python SDK for ISCC generation
- [Amazon Reviews 2023](https://amazon-reviews-2023.github.io/)

## Contact

- **Dataset Issues**: [iscc-datasets GitHub](https://github.com/iscc/iscc-datasets/issues)
- **ISCC Questions**: [ISCC Foundation](https://iscc.io/)