DatasetHelpers

Dataset Processor Tools is a versatile toolkit designed to streamline the processing of image-text datasets for machine learning applications. It empowers users with a range of powerful functionalities to enhance their datasets effortlessly.

Efficiently manage image datasets with Dataset Processor Tools. It supports editing both .txt and .caption files, allowing users to customize and fine-tune dataset annotations easily. Update tags, refine descriptions, and add contextual information with simplicity and seamless organization.

One standout feature is the automatic tag generation using the WD 1.4 SwinV2 Tagger V2 model. This pre-trained model analyzes image content and generates descriptive booru style tags, eliminating the need for manual tagging. Save time and enrich the dataset with detailed image descriptions.

The toolkit also offers advanced content-aware smart cropping. Leveraging the YoloV4 model for object detection, it intelligently identifies images with people and performs automatic cropping. Custom implementation ensures precise cropping, resulting in optimized images ready for further processing. Output dimensions are 512x512, 640x640, or 768x768, compatible with popular machine learning frameworks.