Alibaba Cloud Poses Future Challenge to UK Engineers

2018_2_11_17_48_10

A casual observer may not equate food delivery in China with weather forecasts in the United Kingdom, but a new contest launched by Alibaba Cloud and the UK’s national weather service, the Met Office, does.

Contestants have been tasked with creating algorithms that generate effective delivery routes for anti-gravity balloons in the face of real-time changes in wind speed. The changes, while playing out above ground, are very similar to the real-time shifts in traffic flows faced by Chinese food-delivery companies on the ground.

Called the “Future Challenge,” the contest brings together students and professors from top UK universities, such as Exeter, Oxford and Cambridge, as well as researchers in the private sector. For Alibaba Cloud, it’s a chance to raise its profile in an all-important market.

challenge_proposal_300x167

Contestants in the Future Challenge where tasked with solving this problem.

“Europe boasts some of the best universities in the world, with highly talented professors and students working on complex data science and engineering problems,” said Yeming Wang, deputy general manager of Alibaba Cloud, who is based in London. “One of the aims of this project is to introduce Alibaba Cloud to these great minds, and work together with them on new challenges in the future,” he said.

The Alibaba Group business unit currently has offices in London, Paris and Frankfurt, while customers include Philips and Schneider Electric and Vodafone is a regional partner. Alibaba Cloud also has a data center in Germany.

The challenges faced by food-delivery companies, such as Alibaba Group-owned Ele.me, are real. In China’s hyper-competitive food-delivery space, the faster food arrives, the more likely they’ll retain that customer. To edge out the competition, Alibaba Cloud’s algorithms provide delivery drivers with the fastest routes to their customers in real time. Hence, the contest.

Here’s how it works: The contest is run on Alibaba Cloud’s crowd-intelligence platform, Tianchi. The platform has attracted over 120,000 data scientists from across 77 countries in disciplines ranging from data mining and artificial intelligence to machine learning and operational research. And they have collaborated online to solve real-world issues facing society at large. In a recent competition, 3,600 teams crunched data to find ways to minimize tollgate traffic jams in China’s often-congested cities.

For the Future Challenge, contestants must calculate the best routes for 10 different balloons to destinations across the UK over five days, taking into account wind speed, which could potentially destroy the balloons. The Met Office has delivered multiple wind-speed predictions that contestants use to calculate and approximate a consolidated hourly forecast on which to base their routes.

Alibaba Cloud last month went on a roadshow to three UK universities, Cambridge, Oxford and Exeter, where the Met Office is based, to drum up interest. It also held a hackathon, in London, where engineers from academia and the private sector worked on algorithms in preparation for the larger contest. Three winning teams took home cash prizes.

MetOffice2.png

Wang said Alibaba Cloud wanted to put contestants in the shoes of the Met Office, using multiple predictions based on a large set of data, to work out an actual forecast. He called it a “true challenge,” and one with immediate relevance to the science of weather forecasting.

“While we are some way off balloon-based parcel delivery, it is not unthinkable that short-haul air logistics will happen in the future,” he said.

Moreover, the contest gives these engineers the chance to process a vast amount of data in a relative short space of time, a task that mirrors increasing demands in the market. The Met Office uses a supercomputer to track weather and climate conditions, computing a trillion operations per second and generate hundreds of petabytes of data a day on forecasts. Alibaba’s 11.11 Global Shopping Festival is another example, as it processed $25.3 billion worth of transactions in 24 hours and 325,000 orders a second at its peak.

“From an Alibaba Cloud perspective, this contest and data analytics have relevance across our portfolio,” Wang said. “We have solutions in place to help businesses with processing, visualising, exploring, analysing and integrating datasets in a secure and compliant way.”

The Future Challenge is currently in its semi-final stage, which lasts until Feb. 9. Three winners will be announced later this month and earn a trip to Mobile World Congress, in Barcelona. The top three winners receive $8,000, $5,000 and $3,000, respectively.

### 如何安装 Stable Diffusion 的 Poses 模块 在 Stable Diffusion 中,Poses 模块通常用于生成具有特定姿态的人物图像。以下是关于如何安装和配置该模块的相关说明: #### 1. 准备工作 为了能够顺利运行 Poses 模块,需要先完成 Stable Diffusion 的基本环境搭建以及相关依赖项的安装。可以参考零基础 AI 绘画学习资源中的 PDF 和视频教程来熟悉整个流程[^1]。 #### 2. 下载并加载 Pose 检测模型 Pose 模块的核心在于使用预训练的姿态检测模型(如 OpenPose)。这些模型可以通过以下方式获取: - 如果通过 Google Colab 运行 Stable Diffusion,则可以在后台自动下载所需模型文件[^2]。 - 对于本地部署的情况,需手动访问官方仓库或其他可信源下载对应的 Pose 检测模型,并将其放置到指定目录下。 #### 3. 添加 LoRA 插件支持自定义姿势调整 为了让生成图片更贴合目标人物的动作细节,可引入基于 LoRA 技术开发的扩展功能。具体操作如下: - 打开启动器界面进入 **模型管理**; - 转至 **LoRA 模型 (插件)** 页面; - 使用 “添加模型” 功能导入之前准备好的 Pose 相关 LoRA 文件; - 完成上述设置后记得保存更改并重新启动程序[^3]。 #### 4. 应用 img2img 方法实现姿态迁移 当一切就绪之后就可以利用 `img2img` 工具来进行实际创作了: - 切换到软件内的 `img2img` 选项卡; - 将作为参考素材的照片上传上去; - 输入简洁明了的文字指令告诉算法希望转换为何种风格或者保持哪些特征不变的同时改变其他方面比如动作等等[^4]; 最后执行渲染过程等待成果呈现出来! ```bash # 假设已经克隆好了项目库并且激活虚拟环境的话可以直接运行下面命令测试openpose是否正常工作 cd path/to/stable-diffusion-webui/extensions/sd-webui-openpose/ pip install -r requirements.txt python setup.py develop ``` 以上就是有关于怎样把poses加入到stable diffusion里的全过程讲解啦~
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值