Competition Overview


The competition organized by BenchCouncil adopts a circular approach. The topics of competition come from an industry standard AI benchmark suite---AIBench (Technical Report, Bench18).
Every year, any companies and research institutes can join and organize an competition topic before the deadline (August 1st). The organizer need to pay a fee.
Everyone can register the competition before the registration deadline (September 15).
The competition results will be published on October 1st every year. Then the organizing committee of BenchCouncil will send Bench 19 invitation letter to award-winning candidates.
Every competitors can submit papers to BenchCouncil International Symposium on Benchmarking, Measuring and Optimizing (Bench 19).
The award-winners must submit a paper to Bench 19 conference and give a presentation.
All competitors must submit the codes to BenchCouncil open-source code system---BenchHub. The system can be run on BenchCouncil Testbed.
After the competition results published, the competitors can still participate in the competition for the next-year competition.


Call for Competition Organizers


BenchCouncil provides the competition topics (BigDataBench), competition platform (a new technology testbed) and publicity platform (Xinxiu--a science and education communication tools). The datasets provided by organizers can also be a part of BigDataBench benchmarks. Organizers can provide equipments such as accelerators, and these equipments can be the standard configuration of the testbed by donating.


2019 BenchCouncil International AI System and Algorithm Competition


"2019 BenchCouncil International Artificial Intelligence System Competition" is held on the BenchCouncil New Technology Testbed. The competition topics consist of system competition (on many kinds of chips such as Cambricon, X86, RISC-V and so on) and Intellif algorithm competition. These competitions will gather the global wisdom and make contribution to an autonomous and controllable intelligent computing ecosystem.


Competition Rule


1. Each person can only join one team.
2. Each team can include one to three members and choose one mentor. Each winning team needs to attend the Bench 19', publish related papers and give a report at the meeting.
3. Each team can submit the result up to three times each day.


Communication


We will organize the experts' team to give a presentation in universities and institutes. Xinxiu will be used as the communication tool.


Competition Awards


Special Award (Only one): 100,000 CNY
The First Prize for Every Competition Topic: 30,000 CNY
The Second Prize for Every Competition Topic: 20,000 CNY
The Third Prize for Every Competition Topic: 10,000 CNY

Free HTML5 Bootstrap Template

Competition Topic 1: International AI System Competition based on RISC-V

RISC-V is an open-source instruction set architecture based on RISC principles. As an outstanding representative of open source chip technology in recent years, RISC-V has attracted extensive attention in various fields worldwide. This competition topic focuses on the implementation and optimization of convolutional neural network based image classification task on RISC-V.

The docker image of RISC-V simulator: https://hub.docker.com/r/crva/riscv-qemu

Dataset:
The CIFAR-10 dataset consists of 60000 32x32 colour images in 10 classes, with 6000 images per class. In total, it provides 50000 training images and 10000 test images. The dataset is in binary format. The format of each sample is:
<1 x label><3072 x pixel>
Label is an integer between 0 to 9, indicating the category of the sample. The remained 3072 bytes represent the pixels of RGB channels, which are stored as row-major sequential order. For example, the first 64 bytes are the first two rows of Red channel. The input is the 3072 bytes while the output is the predicted label value. Download the dataset: https://www.cs.toronto.edu/~kriz/cifar-10-binary.tar.gz

More information-Schedule

Registration

Free HTML5 Bootstrap Template

Competition Topic 2: International AI System Competition based on Cambricon Chip

Cambricon is the forerunner of global intelligent chip area. The "Cambricon 1A" processor, which is proposed at 2016, is the first terminal specific processor for AI in the world, and has been widely used in smart phones. This competition topic focuses on the implementation and optimization of convolutional neural network based image classification task on Cambricon chip.

Dataset:
The CIFAR-10 dataset consists of 60000 32x32 colour images in 10 classes, with 6000 images per class. In total, it provides 50000 training images and 10000 test images. The dataset is in binary format. The format of each sample is:
<1 x label><3072 x pixel>
Label is an integer between 0 to 9, indicating the category of the sample. The remained 3072 bytes represent the pixels of RGB channels, which are stored as row-major sequential order. For example, the first 64 bytes are the first two rows of Red channel. The input is the 3072 bytes while the output is the predicted label value. Download the dataset: https://www.cs.toronto.edu/~kriz/cifar-10-binary.tar.gz

More information-Schedule

Registration

Free HTML5 Bootstrap Template

Competition Topic 3:International AI System Competition based on X86 Platform

This competition topic focuses on the implementation and optimization of matrix decomposition based collaborative filtering task on X86 platform.

Dataset:
Movielens dataset constains movie ratings from multiple users, and the information about movie metadata and user attributes. This competition topic involves in partial rating data, including 27,000,000 movie ratings from 280,000 users. The data format (ratings.csv) is as follows:
userId, movieId, rating, timestamp
userId: the user ID
movieId: the movie ID
rating: the rating data
timestamp: the rating time
Download the dataset: https://grouplens.org/datasets/movielens/

More information-Schedule

Registration



Free HTML5 Bootstrap Template

Competition Topic 4:International 3D Face Recognition Algorithm Competition

Shenzhen Intellifusion Technologies Co. Ltd. is an AI (artificial intelligence) unicorn company equipped with AI algorithm, AI chip and big data concurrently for the first time in China.
This competition---3DFRC (3D Face Recognition Challenge) aims at soliciting new approches to advance the state-of-the-art in face recognition.

Dataset:
The test data contains 77715 samples covering 253 face IDs. Each sample includes a RGB image and depth image, with the bit width of 8. This dataset is used to evaluate the models.

More information-Schedule

Registration