ICML Best Paper SD3 Has a Public Tutorial Online! DreamBench++ Is a New Benchmark for Automatic Image Evaluation, Achieving Deep Alignment of Human Preferences
Recently, the best papers of ICML 2024 were announced! Among them is the "Internet celebrity model" of image generation of the year - Stable Diffusion 3 (SD3 for short). SD3 is the latest text-to-image generation model developed by Stability AI, and it has been open sourced on the entire network some time ago!HyperAI Super Neuro has now launched a tutorial on how to run SD3 in ComfyUI workflow.Everyone is welcome to experience the technological innovation of SD3 while reading the paper!
SD3 Tutorial Link:https://go.hyper.ai/ojO3g
From July 22nd to July 26th, hyper.ai official website updates:
* High-quality public datasets: 10
* Selection of high-quality tutorials: 3
* Community article selection: 4 articles
* Popular encyclopedia entries: 5
* Top conferences with deadline in August: 4
Visit the official website:hyper.ai
Selected public datasets
The dataset contains 30 subjects of different categories, including 9 living subjects (such as dogs and cats) and 21 objects, with 4 to 6 images per subject. It allows a model to be trained with a small number of images, enabling it to generate images of that specific individual in many different contexts while maintaining its key visual features.
Direct use:https://go.hyper.ai/Jiqg6
2. ChlD Large-Scale Chinese Idioms Dataset
The dataset contains 581K passages and 729K blanks, and covers multiple fields. In ChID, idioms in passages are replaced by blank symbols. For each blank, a list of candidate idioms including the golden idiom is provided as a choice.
Direct use:https://go.hyper.ai/dt4AR
3. CCPM Chinese Classical Poetry Matching Dataset
This dataset is the Chinese Classical Poetry Matching Dataset launched by Tsinghua University in 2021, which includes a training set (21,778 sentences), a validation set (2,720 sentences) and a test set (2,720 sentences).
Direct use:https://go.hyper.ai/ymhF6
4.MMDU Super-long Multi-image Multi-turn Dialogue Understanding Dataset
The MMDU benchmark consists of 110 high-quality multi-image multi-turn dialogues with more than 1,600 questions, each with a detailed long answer. Questions in MMUD involve 2 to 20 images, with an average image and text token length of 8.2K tokens and a maximum image and text length of 18K tokens, posing a significant challenge to existing multimodal large-scale models.
Direct use:https://go.hyper.ai/vNyjl
5. ModeINet10 Princeton 3D Object Dataset
The ModelNet10 dataset is a part of the ModelNet40 dataset, which contains 4,899 pre-aligned shapes of 10 types of CAD furniture models such as bathtubs, beds, chairs, tables, etc. Among them, 3,991 (80%) shapes are used for training and 908 (20%) shapes are used for testing.
Direct use:https://go.hyper.ai/ZPFKs
The dataset contains an image folder and a label folder. The image folder contains two subfolders train (374 images) for training and Val (111 images) for validation.
Direct use:https://go.hyper.ai/WAKTy
7. baike_qa2019 Encyclopedia Q&A JSON version dataset
The dataset contains 1.5 million pre-filtered, high-quality questions and answers, each question belongs to one category. There are 492 categories in total, of which 434 categories have a frequency of 10 times or more.
Direct use:https://go.hyper.ai/3KWJ8
8. DreamBench++ Image Automatic Evaluation Benchmark Dataset
DreamBench++ is a new benchmark jointly launched in 2024 by researchers from Tsinghua University, Xi'an Jiaotong University, University of Illinois at Urbana-Champaign, Chinese Academy of Sciences, and Megvii, which aims to solve the problems in the evaluation of personalized image generation technology. It introduces the multimodal GPT-4o, achieves deep alignment with human preferences and automated evaluation, and launches a more comprehensive and diverse dataset.
Direct use:https://go.hyper.ai/glVDV
9. COVID-19 Radiography Database Chest X-ray Image Database
The dataset contains 3,616 COVID-19 positive cases, 10,192 normal cases, 6,012 lung opacity (non-COVID-19 lung infection) cases, and 1,345 viral pneumonia images and corresponding lung mask images to help researchers conduct research during the COVID-19 pandemic.
Direct use:https://go.hyper.ai/89Wxz
10. Oceanlnstruct Ocean Large Model Command Dataset
The dataset contains 20,000 instructions and is designed to provide training data for large-scale language models in the ocean field. These instructions cover a wide range of marine scientific knowledge, ensuring that the model has professional capabilities in marine science question answering, content generation, and underwater embodied intelligence.
Direct use:https://go.hyper.ai/WuYlv
For more public datasets, please visit:
Selected Public Tutorials
Stable Diffusion 3 Medium (SD3), an open-source model from Stability AI, has significantly improved image quality, complex prompt understanding, and resource efficiency. It can generate images with realistic details, bright colors, and natural lighting, and can adapt to a variety of styles! This tutorial combines SD3's capabilities with ComfyUI's workflow, so you can start your creative journey right away.
Run online:https://go.hyper.ai/ojO3g
2. Kolors Kuaishou large model demo of text and image
Kolors is a large-scale text-to-image generation model based on latent diffusion developed by the Kuaishou Kolors team. After training on billions of text-image pairs, Kolors has shown significant advantages over open-source and closed-source models in terms of visual quality, complex semantic accuracy, and text rendering of Chinese and English characters. This tutorial does not require any commands to be entered, and you can start image generation immediately with one-click cloning.
Run online:https://go.hyper.ai/ur8q7
3. One-click deployment Mistral-Nemo-Instruct-2407
Mistral-Nemo-Instruct-2407 is a fine-tuned version of the Mistral-Nemo-Base-2407 instruction jointly open-sourced by Mistral AI and NVIDIA. Its performance is significantly better than existing smaller or similar-sized models. Mistral NeMo has 12 billion (12B) parameters and a context window of 128k. Its reasoning, world knowledge, and encoding accuracy are leading in similar scales. This tutorial is a one-click deployment of Mistral-Nemo-Instruct-2407. The relevant environment and dependencies have been installed. You only need to clone it to experience reasoning.
Run online:https://go.hyper.ai/zGkci
Community Articles
In the first episode of the "Meet AI4S" series of live broadcasts, HyperAI was fortunate to invite Ding Jiale, a doctoral student in remote sensing and geographic information systems at Zhejiang University. He gave a simple and in-depth explanation of his research results under the title "Neural Networks Provide New Explanations for Spatial Heterogeneity of Housing Prices". This article is a summary of Dr. Ding's sharing.
View the full report:https://go.hyper.ai/g2fXy
The research team of Bai Xiang and Liu Yuliang from Huazhong University of Science and Technology, in collaboration with the University of Adelaide, Anyang Normal University, and South China University of Technology, used an image-based generative model to train a conditional diffusion model OBSD optimized for oracle bone inscriptions deciphering, providing a novel method for ancient character recognition tasks that are difficult to solve in natural language processing. This article is a detailed interpretation and sharing of the relevant paper.
View the full report:https://go.hyper.ai/fLcZU
Autonomous driving has ushered in a new era of "end-to-end", and high-quality datasets play an important role. In response to this, HyperAI has compiled 10 popular open-source autonomous driving datasets for everyone to collect and use.
View the full report:https://go.hyper.ai/5nj1s
The team led by Zhang Ningyu and Chen Huajun from the School of Computer Science and Technology at Zhejiang University proposed the first large language model in the ocean field, OceanGPT, which can answer questions based on instructions from oceanographers and has gained preliminary embodied intelligence capabilities in ocean engineering. This article is a detailed interpretation and sharing of the relevant paper.
View the full report:https://go.hyper.ai/b6tqu
Popular Encyclopedia Articles
1. Scaling Law
2. Masked Language Modeling (MLM)
3. Data Augmentation
4. Long Short-Term Memory Short-Term Memory
5. Quantum Neural Network
Here are hundreds of AI-related terms compiled to help you understand "artificial intelligence" here:
One-stop tracking of top AI academic conferences:https://go.hyper.ai/event
The above is all the content of this week’s editor’s selection. If you have resources that you want to include on the hyper.ai official website, you are also welcome to leave a message or submit an article to tell us!
See you next week!
About HyperAI
HyperAI (hyper.ai) is the leading artificial intelligence and high-performance computing community in China.We are committed to becoming the infrastructure in the field of data science in China and providing rich and high-quality public resources for domestic developers. So far, we have:
* Provide domestic accelerated download nodes for 1300+ public data sets
* Includes 400+ classic and popular online tutorials
* Interpretation of 100+ AI4Science paper cases
* Support 500+ related terms search
* Hosting the first complete Apache TVM Chinese documentation in China
Visit the official website to start your learning journey: