Katteryna/automotive-tools-instance-segmentation-demo

🚗 Automotive Hand Tools Instance Segmentation Demo

Real-World Computer Vision Dataset

A real-world Computer Vision dataset focused on automotive workshops, hand tools, and equipment.

Dataset Overview

This dataset contains 1,076 original real-world images photographed in automotive workshops and related environments.

All images were personally captured by the dataset creator using an iPhone. The dataset contains no AI-generated or web-scraped images.

Key Statistics

• —📸 1,076 original real-world images
• —🔧 49 object classes
• —🎯 1,160 annotated object instances
• —✏️ 1,160 manually created polygon annotations
• —📍 116,095 polygon points
• —🌍 100+ unique backgrounds and real-world environments
• —✅ 0 invalid annotation lines
• —🏷️ Instance-level polygon segmentation
• —📦 YOLOv8 Segmentation format
• —📦 COCO format available

Main Classes

The dataset includes a variety of automotive workshop tools, equipment, and related objects, including hand tools and objects commonly found in real workshop environments.

The dataset is designed for Computer Vision applications such as:

• —Instance segmentation
• —Object detection
• —Automotive Computer Vision
• —Workshop automation
• —Industrial inspection
• —Robotic perception
• —Tool recognition
• —Object recognition in cluttered environments
• —Real-world perception and AI training

Annotation

All annotations were manually created using polygon-based segmentation.

The annotations are designed to capture object boundaries at instance level rather than providing only bounding boxes.

The dataset was manually reviewed and checked for invalid annotation lines.

Image Characteristics

The images were collected in real automotive workshop environments rather than generated or collected from online image repositories.

The dataset includes variation in:

• —Object position and orientation
• —Lighting conditions
• —Backgrounds
• —Object scale
• —Object combinations
• —Partial occlusions
• —Real-world workshop clutter

This variation makes the dataset suitable for experimenting with Computer Vision models in less controlled environments.

Dataset Formats

The dataset is primarily prepared for:

• —YOLOv8 Segmentation
• —COCO Instance Segmentation

Additional annotation formats can be generated or provided depending on project requirements.

Intended Use

This dataset can be used for research, prototyping, model development, benchmarking, and commercial Computer Vision applications subject to the applicable license.

Potential applications include automotive workshops, industrial automation, robotic perception, visual inspection, tool recognition, and object segmentation.

Data Collection

All photographs were personally captured by the dataset creator in real-world automotive workshop environments.

No AI-generated images or web-scraped images were used.

Dataset Status

Current version: 1,076 images

The dataset may be expanded with additional real-world images and classes based on project requirements.

License

This dataset is distributed under a Commercial License.

Please refer to LICENSE.md for the complete license terms.

Contact

Ekaterina Bocharova

For licensing inquiries, custom dataset development, enterprise licensing, or commercial collaboration, please get in touch.

LinkedIn

www.linkedin.com/in/екатерина-бочарова-0414a1416

Email

kattya.bocharova@gmail.com