File size: 4,623 Bytes
6514d1a
a12710f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6514d1a
 
25f506b
6514d1a
25f506b
6514d1a
25f506b
 
 
 
 
6514d1a
25f506b
 
 
6514d1a
 
 
7ce246e
 
25f506b
7ce246e
 
 
 
6514d1a
 
 
 
 
25f506b
 
6514d1a
a12710f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d5c517c
a12710f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b54ce2d
a12710f
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
---
license: apache-2.0
task_categories:
  - image-classification
  - text-classification
  - zero-shot-classification
tags:
  - e-commerce
  - product-categorization
  - taxonomy
  - multimodal
  - shopify
  - retail
  - product-catalog
size_categories:
  - 10K<n<100K
dataset_info:
  features:
  - name: product_title
    dtype: string
  - name: product_description
    dtype: string
  - name: product_image
    dtype: image
  - name: potential_product_categories
    list: string
  - name: ground_truth_brand
    dtype: string
  - name: ground_truth_is_secondhand
    dtype: bool
  - name: ground_truth_category
    dtype: string
  splits:
  - name: train
    num_bytes: 7425492387
    num_examples: 38631
  - name: test
    num_bytes: 2085920280
    num_examples: 9658
  download_size: 9496974081
  dataset_size: 9511412667
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
---

# The Catalogue: Product Taxonomy Classification Benchmark

A large-scale, multimodal benchmark dataset for product taxonomy classification, featuring real e-commerce products with images, descriptions, and hierarchical category labels.

## Dataset Description

**The Catalogue** is a benchmark dataset designed to evaluate AI models on the task of classifying products into a standardized taxonomy. Each sample includes a product image, title, description, brand, and the ground-truth category from Shopify's product taxonomy.

This dataset is ideal for:
- Evaluating vision-language models on real-world product classification
- Benchmarking multimodal understanding in e-commerce contexts
- Testing hierarchical classification capabilities
- Comparing different approaches to product categorization

### Dataset Statistics

| Metric | Value |
|--------|-------|
| Total samples | 48,289 |
| Unique categories | 10,476 |
| Unique brands | 28,913 |
| Products with descriptions | 92.9% |
| Products with brand | 98.2% |
| Average category depth | 4.5 levels |
| Category depth range | 1-8 levels |

### Top-Level Category Distribution

| Category | Count | Percentage |
|----------|-------|------------|
| Home & Garden | 7,912 | 16.4% |
| Sporting Goods | 6,968 | 14.4% |
| Arts & Entertainment | 5,558 | 11.5% |
| Hardware | 5,137 | 10.6% |
| Vehicles & Parts | 2,555 | 5.3% |
| Business & Industrial | 2,435 | 5.0% |
| Electronics | 2,344 | 4.9% |
| Apparel & Accessories | 2,173 | 4.5% |
| Health & Beauty | 2,162 | 4.5% |
| Food, Beverages & Tobacco | 2,036 | 4.2% |
| Animals & Pet Supplies | 1,922 | 4.0% |
| Furniture | 1,707 | 3.5% |
| Baby & Toddler | 1,226 | 2.5% |
| Toys & Games | 1,153 | 2.4% |
| Office Supplies | 1,129 | 2.3% |
| Cameras & Optics | 967 | 2.0% |
| Other | 900 | 1.9% |

## Dataset Structure

### Data Fields

| Field | Type | Description |
|-------|------|-------------|
| `product_title` | string | Product title |
| `product_description` | string | Product description (may be empty) |
| `product_image` | Image | Product image |
| `potential_product_categories` | list[string] | Candidate category paths |
| `ground_truth_brand` | string | Brand name (may be empty) |
| `ground_truth_is_secondhand` | boolean | Whether the product is secondhand |
| `ground_truth_category` | string | Ground truth category path |

### Category Format

Categories follow a hierarchical path format:
```
Electronics > Computers > Laptops
Home & Garden > Kitchen & Dining > Cookware
```

### Data Splits

| Split | Samples | Percentage |
|-------|---------|------------|
| Train | 38,631 | 80% |
| Test | 9,658 | 20% |

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("Shopify/product-catalogue")

# Access splits
train_data = dataset["train"]
test_data = dataset["test"]

# Example
sample = train_data[0]
print(f"Title: {sample['product_title']}")
print(f"Category: {sample['ground_truth_category']}")
sample['product_image'].show()
```

## Evaluation

The primary task is to predict the `ground_truth_category` given the product's image, title, and optionally description/brand.

Suggested metrics:
- Hierarchical F1 for categories
- Precision and recall for brand and is_secondhand

## Source

Products were sampled from Shopify's merchant catalog, representing real e-commerce products across diverse categories and industries.

## License

Apache 2.0

## Citation

```bibtex
@dataset{product-catalogue,
  title={The Catalogue: Product Taxonomy Classification Benchmark},
  author={Shopify},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/Shopify/the-catalogue-public-beta}
}
```