Problem solved

Building labeled image datasets often involves a tedious cycle of capturing photos, manually sorting them into class folders on a computer, and uploading them to cloud storage. This process is time-consuming, error-prone, and breaks the flow of data collection. TagSet solves this by making the entire workflow instantaneous: capture, label, and upload happen in one step on a mobile device, directly populating a training-ready folder structure in the user's own storage.

Use cases

• Rapid collection of training images for classification models
• Field data gathering for recycling or waste sorting computer vision systems
• On-site dataset creation for industrial quality inspection
• Educational projects where students need to build custom image datasets quickly
• Prototyping computer vision apps without investing in complex annotation tools
• Collecting geographically distributed data for research without central servers

Audience

• ML Engineers
• Data Scientists
• Researchers
• Hobbyists
• Developers

Pros

• Extremely fast capture-to-upload workflow
• No backend; data stays private and secure on your own storage
• Automatically organizes images into training-ready folder structure
• Works with any S3-compatible cloud provider
• Free with no accounts or analytics tracking

Cons

• Currently only available on iOS (Android coming soon)
• Limited to simple class-label tagging; no bounding boxes or segmentation
• Requires manual setup of cloud storage credentials
• No built-in data augmentation or preprocessing features
• Relies on external cloud storage; no local-only mode for offline use