thomasavare/waste-classification-v2

---
language:
- en
size_categories:
- 10K<n<100K
---

# Dataset Card for Dataset Name

## Dataset Description

- **Homepage:**
- **Repository:**
- **Paper:**
- **Leaderboard:**
- **Point of Contact:**

### Dataset Summary

Dataset used to train a language model to do classification on 50 different waste classes.

### Languages

English

## Dataset Structure

### Data Instances

Phrase | Class | Index
-------|-------|-------
"I have this apple phone charger to throw, where should I put it ?" | PHONE CHARGER | 26
"Should I recycle a disposable cup ?" | Plastic Cup | 32
"I have a milk brick" | Tetrapack | 45

### Data Fields

- Phrase
- Class
- Class_index

### Data Splits

train: 12.5K rows

test: 5.38K rows

additional data: 7.24K rows (unseen_phrases.csv)

## Dataset Creation

Manualy with objects and phrases templates.

### Annotations

#### Annotation process

Each object was annotated and then the phrases were annotated according to the object according to its annnotation.

#### Who are the annotators?

Myself

### Personal and Sensitive Information

None

## Considerations for Using the Data

### Social Impact of Dataset

None

### Discussion of Biases

Some classes are more present than others but the dataset is balanced overall. Because it was created using patterns, might not be very robust.

### Other Known Limitations

Repetition of phrase patterns, have to verify performances of model on external phrases for robustness.

原始信息汇总

数据集概述

数据集描述

• 用途: 用于训练语言模型进行50种不同垃圾类别的分类。
• 语言: 英语

数据集结构

数据实例

短语	类别	索引
"I have this apple phone charger to throw, where should I put it ?"	PHONE CHARGER	26
"Should I recycle a disposable cup ?"	Plastic Cup	32
"I have a milk brick"	Tetrapack	45

数据字段

• Phrase: 短语
• Class: 类别
• Class_index: 类别索引

数据分割

• 训练集: 12.5K行
• 测试集: 5.38K行
• 额外数据: 7.24K行 (unseen_phrases.csv)

数据集创建

注释过程

• 对象注释: 每个对象先被注释。
• 短语注释: 根据对象的注释对短语进行注释。
• 注释者: 数据集创建者本人

个人和敏感信息

• 信息状态: 无

使用数据集的考虑因素

数据集的社会影响

• 影响状态: 无

数据集的偏见

• 偏见描述: 某些类别比其他类别更常见，但整体数据集是平衡的。由于使用模式创建，可能不够健壮。

其他已知限制

• 限制描述: 短语模式重复，需要验证模型在未见过的短语上的性能以评估健壮性。