Platform For AI:ArtLab

Background information

With significant progress made in large models, especially AIGC models, a tool that can help developers resolve content generation issues by using AI and large models becomes an urgent need. High requirements of large models for AI technologies and programming raise the costs of learning and using models. To resolve these issues, ArtLab is developed. ArtLab provides an easy-to-use GUI that frees you from focusing on the details of underlying technologies. By using ArtLab, you can complete automated design based on the capabilities provided by AI without writing programs.

Overview

ArtLab is an automated design tool provided by PAI for designers. It is an all-in-one AIGC-based design platform that provides features to support dataset management, model training, AI-based image generation, and online inference. ArtLab integrates with mainstream AIGC large models for text-to-image generation to automate batch image generation. You can use the out-of-the-box features of ArtLab to implement AIGC capabilities in design, art, and creative industries.

ArtLab provides the following features:

Dataset management

• Upload datasets: You can upload your own datasets to ArtLab and add tags to multiple datasets at a time. You can also use datasets as materials for model training.

• Add tags to multiple datasets at a time: Tags can be automatically added to multiple datasets at a time by using Kohya_ss to train the models provided by the system.

Model library

• Models: ArtLab presets mainstream models in the industry. You can add popular models that are displayed on the Models tab to the My Models tab and use them to generate images.

• My Models: You can manage the stored models on the My Models tab. You can add models from the Models tab to the My Models tab or upload models from local directories to the My Models tab.

Toolbox

• ArtLab provides the following preset tools for model training: SDWebUI and Kohya_ss.

Benefits

ArtLab uses the core capabilities of PAI to simplify image generation and improve model quality. This way, you can use intelligent designing services based on your business requirements.

• Optimized data tagging: ArtLab provides easy-to-use features such as uploading, tagging, and using datasets to implement intelligent data tagging with high efficiency and accuracy. This ensures that the training data is fully integrated with your expertise, and therefore builds a quality data foundation for models.

• High fine-tuning efficiency: ArtLab presets mainstream excellent pre-trained models and fine-tunes these models by using the quality data tagging feature. You can use the parameter optimization feature to achieve better model performance by debugging models in an efficient manner.

Trial application

Join one of the following DingTalk groups to apply for a trial of ArtLab:

• ArtLab DingTalk group 3: 84175001108.

• ArtLab DingTalk group 2: 29445042141. This DingTalk group is full.

• ArtLab DingTalk group 1: 47685001864. This DingTalk group is full.