laitimes

The first benchmark dataset for hyperspectral salient object detection was launched

author:HyperAI

This week, OpenAI and Google dropped nuclear bomb-level products in succession at the press conference. OpenAI has always been in the news, releasing GPT-4o ahead of Google's I/O conference. And Google has also made frequent big moves, and it is tough with OpenAI.

In terms of this week's product launches, OpenAI and Google's products each have advantages in the field of artificial intelligence. HyperAI has summarized the essence of the two conferences for you, come and find out!

5 月 13 日-5 月 17 日,hyper.ai 官网更新速览:

* High-quality public datasets: 10

* A selection of high-quality tutorials: 3

* Selected community articles: 4

* Popular Encyclopedia Entries: 5 entries

Visit the official website: hyper.ai

Curated public datasets

1. DVQA Image Q&A Dataset

DVQA is a dataset that tests bar chart understanding in a Q&A framework that covers a large number of generic histogram styles, containing approximately 3.48 million question-answer pairs for 300,000 images.

Direct use: https://go.hyper.ai/p4183

2. Exoplanets 系外行星数据集

The dataset contains information on 4,856 exoplanets like name, mass, year of discovery, distance from Earth, and more.

Direct use: https://go.hyper.ai/OcA3h

3. Safebooru 动漫图像元数据集

Safebooru is a tag-based image archive maintained by anime enthusiasts that contains 3 million lines of tag-based anime image metadata. It allows users to post images and add tags, annotations, translations, and comments.

Direct use: https://go.hyper.ai/IyCRb

4. Marvel Universe Social Network Dataset

The dataset contains Marvel heroes, comics, and the relationships between them. Can be used to study social networks in Marvel hero comics.

Direct use: https://go.hyper.ai/9GQuu

5. Marvel Comic Books 漫威漫画数据库

The dataset contains information about all the comic books that have ever been published in the Marvel Universe, such as name, author, active year, and price.

Direct use: https://go.hyper.ai/dRBXp

6. Star Wars Movie Script Dataset

Star Wars is a popular movie series set in a galaxy far, far away, with a total of 9 films in the series. The dataset contains a collection of scripted dialogues between movie characters from 4-6 that can be used for text mining and linguistics.

Direct use: https://go.hyper.ai/7g25U

7. WIE World Events Dataset

The dataset spans important historical milestones from antiquity to modern times, covering a variety of global events. It provides a comprehensive timeline of the events that shaped the world, providing insights into wars, cultural shifts, technological advances, and social movements.

Direct use: https://go.hyper.ai/Dw3tT

8. HSOD-BIT-V1 Hyperspectral Salient Object Detection Benchmark Dataset

HSOD-BIT is the first large-scale, high-quality benchmark dataset for hyperspectral salient object detection, designed to leverage the advantages of spectral information to achieve greater accuracy in salient object detection tasks. The dataset provides pixel-level manual annotation for 319 hyperspectral data cubes and generates corresponding pseudo-color images for the data needs of contemporary deep learning models.

Direct use: https://go.hyper.ai/HGnjm

9. AFHQ Animal Face Image Dataset

The dataset consists of 16,130 high-quality images at 512×512 resolution and includes images of animal faces in three categories: cats, dogs, and wildlife.

Direct use: https://go.hyper.ai/AbByr

10. WikiArt Faces 艺术面孔数据集

The dataset contains hundreds of face images in 26 styles, selected from WikiArt works, with varying degrees of realism and stylization. The size or aspect ratio of the image is inconsistent, so for most applications, some preprocessing (filling, scaling, etc.) is required.

Direct use: https://go.hyper.ai/OO5RX

For more public datasets, please visit:

https://hyper.ai/datasets

A selection of public tutorials

1. Online tutorial|Turing Award winner Yann LeCun praises! Xiaohongshu open-source InstantID, one original picture can be customized in a variety of styles

InstantID is a diffusion-based image generation solution that enables rapid generation from a single reference image to a variety of stylized portraits. Just upload a selfie and get a customized AI photo in 20 seconds.

Run online: https://go.hyper.ai/Eq4zl

2. ComfyUI InstantID 工作流在线教程

The ComfyUI InstantID workflow can transform an ordinary portrait photo into a work of art with a highly personalized style, and in the process of style transfer, the algorithm will combine the content of one image with the style of another image to generate a new image that retains the content of the original image and has the artistic style of another image. This technique is widely used in fields such as art creation, image processing, and visual effects.

Run online: https://go.hyper.ai/djKs3

3. Preprocess audio data using the GTZAN dataset

The tutorial will use the GTZAN dataset to explain and demonstrate all the possible features extracted from the audio data that will help with model training.

Run online: https://go.hyper.ai/9Teed

We have also established a Stable Diffusion tutorial exchange group, welcome friends to join the group to discuss various technical issues and share application effects~

A selection of community articles

1. GPT-4o Dragon Slayer Shock! Multi-modal, real-time interaction, free for all, silky voice interaction ChatGPT has won the game

OpenAI released its latest flagship large model, GPT-4o, on Tuesday, which is free for all users to use! The model provides GPT-4-level AI capabilities, accepts any combination of text, audio, and image inputs, and is capable of generating any combination of text, audio, and image outputs. Based on GPT-4o, the upgraded ChatGPT supports real-time interaction, with a reaction time similar to that of human conversations. For more first-hand reports and model effect displays, please refer to this article.

Get detailed dry goods: https://go.hyper.ai/7SZEG

2. The era of Google Gemini is here! Strengthened the search moat and empowered the whole family bucket, Gemini 1.5 Pro was upgraded to 2 million tokens

This article summarizes several highlights of the Google I/O conference: such as the upgrade of Gemini 1.5 Pro to 2 million tokens, the launch of the lightweight model Gemini 1.5 Flash, and the integration of AI Overviews into Google Search, which provides an overview of information and assists users in life planning. More first-hand reports are detailed in this article.

See the full story: https://go.hyper.ai/vqptC

3. Get up to $300 cash! Share hands-on experience and you'll be rewarded with ease

The second phase of the "Creator Incentive Program" of OpenBayes Computing Platform is online, come and participate with HyperAI! Whether you're a veteran of deep learning, data science, or large models, or you're just getting started; Whether you're a veteran with a lot of creative experience or just starting out in the world of social media; As long as you are willing to share, you will be rewarded, and the chance of winning is 100%!

Full event information: https://go.hyper.ai/3uBMM

4. Don't be afraid of space matter "hitting the earth"! The Chinese Academy of Sciences team proposed a new method for coronal mass ejection identification, which is close to the human identification results

The research team from the Key Laboratory of Solar Activity and Space Weather of the National Space Science Center of the Chinese Academy of Sciences proposed a method for CME identification and parameter acquisition based on machine learning. Compared with the CME manual catalog and some classic automatic identification methods, this method has high efficiency and speed, which can identify relatively weak CME signals and give accurate CME morphological information. This article is a detailed interpretation and sharing of the study.

See the full story: https://go.hyper.ai/0PFQD

A selection of popular encyclopedia entries

1. 召回率 Recall Rate

2. 掩码语言建模 MLM

3. Extended long short-term memory xLSTM

4. Kolmogorov-Arnold 表示定理

5. Kolmogorov-Arnold Networks

Here are hundreds of AI-related entries, so you can read "artificial intelligence" here:

https://hyper.ai/wiki

Station B live broadcast preview

Apple will host WWDC 2024 from June 10-14, and to help you learn more about Apple, the Super Nerve B live broadcast room will continue to broadcast "Apple Featured" videos from next week, covering past WWDC conferences, executive interviews, related documentaries and more.

The following table is a preview of the content selected by the editor for everyone↓↓↓

date Time substance

May 20th

Monday

18:00 Biography of Steve Jobs

May 21

Tuesday

18:00 Interview with Steve Jobs 1990

May 22nd

Wednesday

18:00

Interview

Jobs vs Bill Gates

May 23rd

Thursday

18:00

The iPhone was first introduced

Conference

May 24th

Friday

18:00 History of Steve Jobs

May 25th

Saturday

18:00

How is Apple doing

Survived almost bankruptcy

May 26th

Sunday

18:00 History of Tim Cook

Super Neural TV broadcasts live 24×7 hours non-stop, and you can harvest the "electronic mustard" in the field of AI with a click:

http://live.bilibili.com/26483094

That's all for this week's Editor's Picks, if you have resources that you want to include on the hyper.ai official website, you are welcome to leave a message or contribute to let us know!

See you next week!

About HyperAI (hyper.ai)

HyperAI (hyper.ai) is a leading artificial intelligence and high-performance computing community in China, committed to becoming the infrastructure of the domestic data science field, providing rich and high-quality public resources for domestic developers.

* Provide domestic accelerated download nodes for 1200+ public datasets

* Includes 300+ classic and popular online tutorials

* 解读 100+ AI4Science 论文案例

* Support 500+ related terms query

* Hosted the first complete Apache TVM Chinese documentation in China

Visit the official website to start your learning journey:

https://hyper.ai/

Read on