DataBloom

Data is one of the most valuable assets that a business can possess. It sits at the core of data science and data analysis: without data, they’re both…

Data is one of the most valuable assets that a business can possess. It sits at the core of data science and data analysis: without data, they’re both obsolete. Businesses that actively collect data may have a competitive advantage over those that do not. With sufficient data, organizations can better determine the cause of problems and make informed decisions.

There are scenarios where an organization may lack sufficient data to draw necessary insights. For example, a start-up almost always begins with no data. Instead of moping about their deficiencies, a better solution is to employ data acquisition techniques to help build a custom database.

This post covers a popular data acquisition technique called web scraping. You can follow along using the code in the kurtispykes/web-scraping-real-estate-data GitHub repository.

What is data acquisition?

Data acquisition (also referred to as DAQ) may be as simple as a technician logging the temperature of your oven. You can define DAQ as the process of sampling signals that measure real-world physical phenomena and converting the resulting samples into digital numerical values that a computer can interpret.

In an ideal world, we would have all data handed to us, ready for use, whenever we wanted. However, the world is far from ideal. Data acquisition techniques exist because some problems require specific data that you may not have access to at a particular time. Before any data analysis can be conducted, the data team must have sufficient data available. One technique to acquire data is web scraping.

What is web scraping?

Web scraping is a popular data acquisition technique that has become a hot topic of discussion among those with rising demands for big data. Essentially, it’s the process of extracting information from the Internet and formatting it to be easily usable in data analytics and data science pipelines.

In the past, web scraping was a manual process. The process was tedious and time-consuming, and humans are prone to error. The most common solution is to automate. Automation of web scraping enables you to speed up the process while saving money and reducing the likelihood of human error.

However, web scraping has its challenges.

Challenges of web scraping

Building your own web scraper has challenges outside of knowing how to program and understanding HTML. It is beneficial to know in advance the various obstacles that you may encounter while data scraping. Here are a few of the most common challenges that you’ll face when scraping data from the web.

robots.txt

Permissions for scraping data are usually held in a robots.txt file. This file is used to inform crawlers about the URLs that can be accessed on a website. It prevents the site from being overloaded with requests.

The first thing to check before you begin a web scraping project is whether the target website permits web scraping. Websites can decide whether to allow web scrapers on their website for web scraping purposes.

Some websites do not permit automated web scraping, typically to prevent competitors from gaining a competitive advantage and draining the server resources from the target site. It does affect the website’s performance.

You can check the robots.txt file of a website by appending /robots.txt to the domain name. For example, check Twitter’s robots.txt as follows: www.twitter.com/robots.txt.

Structural changes

UI and UX developers periodically add, remove, and undergo regular structural changes to a website to keep it up to date with the latest advancements. Web scrapers depend on the code elements of the web page at the time that the scraper is built. Thus, frequent changes to a website may result in data being lost. It’s always a good idea to keep tabs on the changes to a web page.

Also, consider that different web page designers may have different criteria for designing their pages. This means that if you plan on scraping multiple websites, you might have to build multiple scrapers, one for each website.

IP blockers, or getting banned

It is possible to be banned from a website. If your web scraper is sending an unnaturally high number of requests to a website, it’s possible for your IP address to get banned. Alternatively, the website may restrict its access to break down the scraping process.

There’s a thin line between what is considered ethical and unethical web scraping: crossing the line quickly leads to IP blocking.

CAPTCHAs

A CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) does exactly what it says in its name: distinguishes humans from bots. The problems posed by CAPTCHAs are typically logical and straightforward for a human to solve but challenging for a bot to accomplish the same feat, which prevents websites from being spammed.

There are ethical workarounds for scraping websites with CAPTCHAs. However, that discussion is beyond the scope of this post.

Honeypot traps

In a similar fashion to how you would set up devices or enclosures to catch pests that invade your home, website owners set up honeypot traps to catch scrapers. These traps are typically links that are not visible by humans but are visible to web scrapers.

The intention is to get information about the scraper, such as its IP address, so they can block the scraper’s access to the website.

Real-time data scraping

There are certain scenarios where you may need data to be scrapped in real time (for example, price comparisons). As changes can occur anytime, the scraper must constantly monitor the website and scrape data. Acquiring large amounts of data in real time is challenging.

Web scraping best practices

Now that you’re aware of the challenges you may face, it’s important to know the best practices to ensure that you are ethically scraping web data.

Respect robots.txt

One of the challenges that you’ll face when scraping web data is abiding by the terms of robots.txt. It is a best practice to follow the guides set by a website around what a web scrape can and cannot crawl.

If a website does not permit web scraping, it is unethical to scrape that website. It’s better to find another data source or contact the website owner directly and discuss a solution.

Be nice to servers

Web servers can only take so much. Exceeding a web server’s load results in the server crashing. Consider what may be an acceptable frequency of requests to make to a host’s server. Several requests in a short time span may result in server failure, which in turn disrupts the user experience for other visitors to the website.

Maintain a reasonable time lapse between requests and be considerate with the number of parallel requests.

Scrape during off-peak hours

Understand when a website may receive its most traffic and refrain from scraping during these hours. Your goal is not to hamper the user experience of other visitors. It’s your moral responsibility to scrape when a website receives less traffic. (This also works in your favor as it significantly improves the speed of your scraper.)

Scrapy tutorial

Web scraping in Python usually involves coding several menial tasks from scratch. However, Scrapy, an open-source web crawling framework, deals with several of the common start-up requirements by default. This means that you can focus on extracting the data that you need from the target websites.

To demonstrate the power of Scrapy, you develop a spider, which is a Scrapy class where you define the behavior of your web scraper. Use this spider to scrape all the listings from the Boston Realty Advisors website (Figure 1).

Inspecting the target website

Before starting any web scraping project, it is important to inspect the target website to scrape. The first thing that I like to check when examining a website is whether the pages are static or generated by JavaScript. 

To bring up the developer toolbox, press F12. On the Network tab, make sure that Disable cache is checked.

To bring up the command palette, press CTRL + SHIFT + P (Windows Linux) or Command + SHIFT + P (Mac). Type Disable JavaScript, then press Enter and reload the target website.

Figure 2 shows how the Boston Realty Advisors website looks without JavaScript. 

The empty page tells you that the target web page is generated using JavaScript. You can’t scrape the page by trying to parse the HTML elements displayed in the Elements tab of the developer toolbox. Re-enable JavaScript in the developer tools and reload the page.

There’s more to examine.

In the developer toolbox, choose the XHR (XML HTTP Request) tab. After browsing the requests sent to the server, I noticed something interesting. There are two requests called “inventory,” but in one, you have access to all the data on the first page in JSON.

This information is great because it means that you don’t have to visit the main listing page on the Boston Realty Advisors website to get access to the data you want.

Now you are ready to begin scraping. (I did more inspections to better understand how to mimic the requests being made to the server, but that is beyond the scope of this post.)

Creating the Scrapy project

To set up the Scrapy project, first install scrapy. I recommend doing this step in a virtual environment.

pip install scrapy

After the virtual environment is activated, enter the following command:

scrapy startproject bradvisors

This command creates a Scrapy project called bradvisors. Scrapy also automatically adds some files to the directory.

After running the command, the final directory structure looks like the following tree:

.
└── bradvisors
   ├── bradvisors
   │   ├── __init__.py
   │   ├── items.py
   │   ├── middlewares.py
   │   ├── pipelines.py
   │   ├── settings.py
   │   └── spiders
   │       └── __init__.py
   └── scrapy.cfg

So far, you’ve inspected the elements of the website to scrape and created the Scrapy project.

Building the spider

The spider module must be built in the bradvisors/bradvisors/spiders directory. The name of my spider script is bradvisors_spider.py but you can use a custom name.

The following code extracts the data from this website. The code example only runs successfully when the items.py file is updated. For more information, see the explanation after the example.

import json

import scrapy
from bradvisors.items import BradvisorsItem

class BradvisorsSpider(scrapy.Spider):
    name = "bradvisors"
    start_urls = ["https://bradvisors.com/listings/"]
  
    url = "https://buildout.com/plugins/5339d012fdb9c122b1ab2f0ed59a55ac0327fd5f/inventory"

    headers = {
    'authority': 'buildout.com',
    'accept': 'application/json, text/javascript, */*; q=0.01',
    'accept-language': 'en-GB,en-US;q=0.9,en;q=0.8',
    'cache-control': 'no-cache',
    'content-type': 'application/x-www-form-urlencoded; charset=UTF-8',
    'origin': 'https://buildout.com',
    'pragma': 'no-cache',
    'referer': 'https://buildout.com/plugins/5339d012fdb9c122b1ab2f0ed59a55ac0327fd5f/bradvisors.com/inventory/?pluginId=0&iframe=true&embedded=true&cacheSearch=true&=undefined',
    'sec-ch-ua': '"Google Chrome";v="105", "Not)A;Brand";v="8", "Chromium";v="105"',
    'sec-ch-ua-mobile': '?0',
    'sec-ch-ua-platform': '"Windows"',
    'sec-fetch-dest': 'empty',
    'sec-fetch-mode': 'cors',
    'sec-fetch-site': 'same-origin',
    'user-agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/105.0.0.0 Safari/537.36',
    'x-newrelic-id': 'Vg4GU1RRGwIJUVJUAwY=',
    'x-requested-with': 'XMLHttpRequest'
    }
  
    def parse(self, response):
        url = "https://buildout.com/plugins/5339d012fdb9c122b1ab2f0ed59a55ac0327fd5f/inventory"
        # There are 5 pages on the website
        for i in range(5):
            # Change the page number in the payload
            payload = f"utf8=%E2%9C%93&polygon_geojson=&lat_min=&lat_max=&lng_min=&lng_max=&mobile_lat_min=
        &mobile_lat_max=&mobile_lng_min=&mobile_lng_max=&page={str(i)}&map_display_limit=500&map_type=roadmap
            &custom_map_marker_url=%2F%2Fs3.amazonaws.com%2Fbuildout-production%2Fbrandings%2F7242%2Fprofile_photo
                %2Fsmall.png%3F1607371909&use_marker_clusterer=true&placesAutoComplete=&q%5Btype_use_offset_eq_any%5D%5B%5D=
                    &q%5Bsale_or_lease_eq%5D=&q%5Bbuilding_size_sf_gteq%5D=&q%5Bbuilding_size_sf_lteq%5D=&q%5B
                        listings_data_max_space_available_on_market_gteq%5D=&q%5Blistings_data_min_space_available_on_market_lteq
                            %5D=&q%5Bproperty_research_property_year_built_gteq%5D=&q%5Bproperty_research_property_year_built_lteq
                                %5D=&q%5Bproperty_use_id_eq_any%5D%5B%5D=&q%5Bcompany_office_id_eq_any%5D%5B%5D=&q%5Bs%5D%5B%5D="
            # Crawl the data, given the payload
            yield scrapy.Request(method="POST", body=payload, url=url, headers=self.headers, callback=self.parse_api)

    def parse_api(self, response):
        # Response is json, use loads to convert it into Python dictionary
        data = json.loads(response.body)

        # Our item object defined in items.py
        item = BradvisorsItem()

        for listing in data["inventory"]:
            item["address"] = listing["address_one_line"]
            item["city"] = listing["city"]
            item["city_state"] = listing["city_state"]
            item["zip"] = listing["zip"]
            item["description"] = listing["description"]
            item["size_summary"] = listing["size_summary"]
            item["item_url"] = listing["show_link"]
            item["property_sub_type_name"] = listing["property_sub_type_name"]
            item["sale"] = listing["sale"]
            item["sublease"] = listing["sublease"]
            yield item

The code accomplishes the following tasks:

1. Defines the name of the scraper as bradvisors.
2. Defines the headers to pass with the request.
3. Specifies that the parse method is the automatic call back when the scraper is run.
4. Defines the parse method, where you iterate through the number of pages to scrape, pass the page number to the payload, and yield that request. This calls back the parse_api method on each iteration.
5. Defines the parse_api method and converts the valid JSON response into a Python dictionary.
6. Defines the BradvisorsItem class in items.py (next code example).
7. Loops through all the listings in the inventory and scrapes specific elements.

# items.py
import scrapy

class BradvisorsItem(scrapy.Item):
   # define the fields for your item here like:
   # name = scrapy.Field()
   address = scrapy.Field()
   city = scrapy.Field()
   city_state = scrapy.Field()
   zip = scrapy.Field()
   description = scrapy.Field()
   size_summary = scrapy.Field()
   item_url = scrapy.Field()
   property_sub_type_name = scrapy.Field()
   sale = scrapy.Field()

Next, you must execute the scrape to parse the data.

Running the scraper

Navigate to the project’s root directory from the command line (in this case, that is bradvisors). Run the following command:

scrapy crawl bradvisors -o data.csv

This command scrapes the Boston Realty Advisors website and saves the extracted data in a data.csv file in the project’s root directory.

Great, you have now acquired real estate data!

What’s next?

As the demand for big data grows, equipping yourself with the ability to acquire data with a web scraping tool is an extremely valuable skill set. Scraping data from the Internet can present several challenges. As well as acquiring the data that you require, your goal should be to treat websites respectfully and scrape them ethically. 

Did you find this Scrapy tutorial helpful? Leave your feedback in the comments or connect with me:

• /kurtispykes on LinkedIn
• @KurtisPykes on Twitter
• /kurtispykes on GitHub

Web scraping FAQ

Is web scraping illegal?

No, web scraping is legal, given that the data you are scraping is public. Search engines such as Google scrape web data daily to curate search results for their users.

Is web scraping free?

You can pay for web scraping services to simplify the web scraping process. You could also learn a programming language and do it yourself for free.

What tools can I use for web scraping in Python?

There are several tools for web scraping in Python, such as Beautiful Soup, MechanicalSoup, Requests (Python module), Scrapy, Selenium, and Urllib.

A smart hospital relies on data-driven insights, including machine learning models and AI-powered medical devices, to facilitate decision-making.

A smart hospital relies on data-driven insights, including machine learning models and AI-powered medical devices, to facilitate decision-making.

Software teams comprise a broad range of professionals, from software engineers and data scientists to project managers and technical writers. Sharing code with…

Software teams comprise a broad range of professionals, from software engineers and data scientists to project managers and technical writers. Sharing code with other team members is common when working on a project, and it is important to track all changes. This is where pull requests come in.

In software development, a pull request is used to push local changes into a shared repository (Figure 1). It is a way for you to request code review from other collaborators before pushing an approved update to the central server. This helps maintain version control.

This post discusses the benefits of pull requests and shares tips for creating and handling pull requests when working on software projects. Using this information, you will be better equipped to work with many collaborators on major projects.

The steps of a pull request

To create a pull request, follow the steps outlined below.

1. Create a new git branch to work locally using the following command:
   git -b BRANCH_NAME
2. Implement changes and push them frequently (so that they do not get lost) using the following command:
   git add NAME_OF_THE_FILE
git commit -m "DESCRIBE YOUR RECENT CHANGES"
3. After you have finished the implementation and committed your changes locally, you should get the latest changes from the shared repository to ensure there are no conflicting changes. You can get the latest changes using the following command:
   git pull origin BRANCH_NAME
4. Push your changes to the remote repository using the following command:
   git push --set-upstream-to origin REMOTE_BRANCH_NAME
5. Navigate to the user interface of the platform where your shared repository is located (GitLab, GitHub, BitBucket). There you are asked to write the name of the pull request and a short description. You also have the option to assign it to someone from your team to review it.

For a more detailed introduction to pull requests, see Making a Pull Request.

Key benefits of using pull requests 

Whether you are working on the frontend or backend of a project, pull requests can help with the code review process when working with a team. This section details the key benefits of using pull requests in your work.

Facilitate collaboration 

When it comes to collaboration, there are a few things that can make or break a team. One of those things is the ability to work together, even if members are responsible for different parts of the project.

Using pull requests, changes can be made without impacting the work of others. They are a great way to gather tips or code improvements from team members.

If you are unsure about a code change, submit a pull request for feedback. Other team members may have suggestions that you had not considered, and this can help you make better decisions about your code.

In any project, it is important to have experienced engineers review and accept or reject changes since you may miss some things that they can see from a fresh perspective. 

However, it is equally important to avoid bottlenecks when multiple team members are submitting changes to the project codebase. For those working on pull requests, it is critical to set expectations for expected review times. This ensures the project continues to move forward. 

Build features faster

Pull requests are a powerful tool that can help teams build features faster. Because pull requests can be reviewed with comments added, they provide an excellent way to communicate code changes.

First, they enable developers to submit changes to a project without having to wait for the project maintainer to merge the changes. This enables team members to work on code changes in parallel, which can speed up development.

Second, pull requests can be reviewed and comments can be added. Developers reviewing pull requests may need to ask questions or clarify potential errors. You can also use comments to share resources. 

Third, pull requests can be merged, so that changes can be integrated into the project quickly and easily when building new features.

Reduce risks associated with adding new code

There is no doubt that programming code comes with a degree of risk. After all, every time you add something new to your codebase, you are potentially introducing new bugs and vulnerabilities that affect the end user. 

Before a pull request is merged into the main codebase, other team members have the opportunity to review the changes to ensure compliance with the team’s coding standards. Bugs and errors can be addressed before they cause any problems in the live code.

With pull requests, you can always roll back to a previous version in case things go wrong. Pull requests become your safety net.

Improve code quality and performance

When you create a pull request, you are essentially asking for someone else to review your code and give feedback. By engaging a colleague, you can improve the quality of your code based on that feedback.

You can help reviewers understand your changes by writing descriptive commit messages and explanations in the description section of the pull request. 

You can also avoid potential problems if you make a change that someone else does not agree with. They can simply raise an issue with your pull request. This gives you the opportunity to fix the problem before it becomes a bigger issue. This is a powerful way to improve the quality of your code. 

Takeaways

Maintaining version control through pull requests is important for software teams. This approach enables team members to collaborate while tracking and managing changes to software systems. By using pull requests, teams can work on different parts of a system at the same time and then easily merge their changes together. This boosts team efficiency and prevents conflicts.

When used correctly, pull requests provide a clear and concise way to view changes that have been made to the code or file, facilitating discussion and feedback.

The importance of pull requests cannot be overstated. They are an essential part of the software development process, helping to ensure that relevant parties review code changes before they are merged into the main codebase. This helps to avoid bugs and other problems that could potentially cause serious issues. 

Learn to train an end-to-end TTS system and track experiments in this live workshop on December 8. Set up the environment, review code blocks, test the model,…

Learn to train an end-to-end TTS system and track experiments in this live workshop on December 8. Set up the environment, review code blocks, test the model, and more.

Machine learning (ML) security is a new discipline focused on the security of machine learning systems and the data they are built upon. It exists at the…

Machine learning (ML) security is a new discipline focused on the security of machine learning systems and the data they are built upon. It exists at the intersection of the information security and data science domains. 

While the state-of-the-art moves forward, there is no clear onboarding and learning path for securing and testing machine learning systems. How, then, should interested practitioners begin developing machine learning security skills? You could read related articles on arXiv, but how about practical steps?

Competitions offer one promising opportunity. NVIDIA recently helped run an innovative ML security competition at the DEF CON 30 hacking and security conference. Hosted by AI Village, the competition drew more than 3,000 participants. It aimed to introduce attendees to each other and to the field of ML security. The competition proved to be a valuable opportunity for participants to develop and improve their machine learning security skills.

NVIDIA AI Red Team and AI Village

To proactively test and assess the security of NVIDIA machine learning offerings, the NVIDIA AI Red Team has been expanding. While this team consists of experienced security and data professionals, they recognized a need to develop ML security talent across the industry. With more exposure and education, data and security practitioners are likely to improve the security of their deployed machine learning systems.

AI Village is a community of data scientists and hackers working to educate on the topic of artificial intelligence (AI) in security and privacy. The community holds events at DEF CON each year. 

The NVIDIA AI Red Team and AI Village joined together at DEF CON 30 to engage the information security community with a machine learning security competition. The topic was potentially new to many attendees. Members of the AI Village created challenges designed to teach and test elements of ML security knowledge. In addition to NVIDIA, these members represented AWS Security, Orang Labs, and NetSec Explained. 

AI Village Capture the Flag Competition

Capture the Flag (CTF) competitions include multiple challenges. Competitors play through the challenges and collect flags for those that are completed successfully. These flags are assigned various point values based on the level of challenge. Competitors win by collecting the most points. 

With this familiar format in mind, the AI Village and NVIDIA AI Red Team built The AI Village CTF @ DEFCON. Organizers partnered with Kaggle to use a platform familiar to the machine learning community. Similar to information security CTFs, Kaggle competitions provide a format for ML researchers to compete on discrete problems. 

Partnering with Kaggle provided the competition with a flexible and scalable platform that paired compute and data hosting with documentation and scoring. Although the challenge servers are no longer active, you can view the challenge descriptions.

Competitors reported onboarding and moving through the challenges with ease, with minimal additional infrastructure required from the AI Village. Furthermore, Kaggle has a large audience of skilled data scientists and machine learning engineers who were excited to explore the security domain. Kaggle also generously offered ongoing support and $25,000 in prizes. We could not have asked for a better partner for this event.

Over the month-long competition, over 3,000 competitors hacked their way through 22 challenges. This far exceeded expectations and included participants from over 70 countries, from first-time Kagglers to Grandmasters. The event succeeded in bringing the traditional information security and machine learning communities together to tackle a range of challenges from this new domain of ML security. 

Competitors used publicly available tools and innovative technique applications such as open source research, masking, and dimensionality reduction. In the process, they often reimplemented attacks from academic literature as well. 

Because one challenge remained unsolved, there was always a chance for someone to rise to the top of the leaderboard. For the final two weeks of the competition, the Kaggle Discussion Board and AI Village Discord were abuzz with theories and explorations of the remaining unsolved challenge. The organizers were checking hourly for a buzzer-beating leaderboard shift. Check out the challenge solutions.

Inference Challenge

In the Inference Challenge, participants had to execute a membership inference attack to identify training samples. They only had API access to an image classifier. When done successfully, the competitors would identify images that showed characters of the flag. 

Some competitors chose to randomly generate the images by permuting pixel values, effectively brute-forcing the problem. Other competitors assumed that the training data may have included a standard dataset and used EMNIST as their source data, leveraging open source data. Others made use of the Adversarial Robustness Toolbox, producing output similar to what is shown in Figure 2. 

Whatever method used, a successful challenger would be rewarded with the flag, spelling D3FC0N. This leetspeak encoding of the DEF CON conference name was used in several places on the conference website. 

Crop2 Challenge

Research and out-of-the-box thinking often help to solve CTF challenges. For instance, the one unsolved challenge in the competition was Crop2. In Crop2, participants were given a poisoned cropping model and had to create the poisoned sample (within some error bounds). They had one training data example to work with (Figure 3). 

This is a difficult problem without an efficient, standard algorithmic solution. When you think about all of the pixels in an image and all of the possible pixel values across three color channels, the search space explodes to over 800 billion options. Instead, competitors could combine reverse engineering, open source research, and assumptions to reduce the number of combinations.

After the competition ended, organizers gave hints to help competitors solve the Crop2 Challenge. Some of the key hints included using open source research to determine that pixel colors likely were generated by matplotlib default colormaps. This greatly reduces the search space into the hundreds of thousands. 

By making these informed assumptions, one competitor was eventually able to reach the Crop2 Challenge solution. One trait of great hackers is tenacity: still working tirelessly after the competition ended, this competitor diligently worked through the provided hints. The competitor reported that a hint “helped me realize that we only needed to use nine colors. Mate, I’d been fiddling around with 16 million. This made the search space manageable.” 

Competitor notebooks

Check out some of our favorite notebooks from competitors:

1. Chris Deotte – From a member of the Kaggle Grandmasters of NVIDIA (KGMoN), these solutions are very well organized and documented. We recommend Secret Sloth in particular.
2. Eric Bouteillon – Watch the flag appear character-by-character in Excuse Me. Also notice the different solve techniques for the MATH challenges. Have you heard of silhouette score?
3. John MacGillivray – John deduced that the Hotterdog model was based on MobileNet, enabling an offline attack. Great tradecraft.
4. Fournierp – A comprehensive writeup about model inversion for the Inference Challenge, written from scratch. You can also check out the MIFace in the Adversarial Robustness Toolbox.
5. Eoin O – Learn how you could have solved the Crop2 Challenge. More than 3,000 competitors tried to solve it for the greater part of a month. The day after the competition ended, organizers released several hints. Within a few hours, it was solved. It was great to see all of the competitors collaborating in Discord and the Kaggle Discussion Board after the competition ended.

Summary

The AI Village CTF @ DEF CON 30 competition showed that there is a significant appetite in both the security and data professions to improve machine learning security skills. As ML systems are deployed in increasingly security-critical contexts, it will become imperative to train professionals and develop tools and methods for security development, deployment, and testing. 

NVIDIA will continue driving innovation with a robust and secure ecosystem for AI, from embedded devices and laptops to supercomputers and the cloud. As part of this effort, our AI Red Team will empower ML security research and testing internally and establish security practices across the industry. We will host competitions, workshops, and release research and security tools in the future. If you’re interested in participating, contact us at threatops@nvidia.com.

Additional resources

• Learning to Defend AI Deployments Using an Exploit Simulation Environment
• Introduction to Adversarial Robustness
• Adversarial Machine Learning Reading List
• HackingNeuralNetworks GitHub repo 
• Coefficient GitHub repo 
• HackThisAI GitHub repo

Denoising is an advanced technique used to decrease grainy spots and discoloration in images while minimizing the loss of quality.

Denoising is an advanced technique used to decrease grainy spots and discoloration in images while minimizing the loss of quality.

Self-driving cars must be able to detect objects quickly and accurately to ensure the safety of their drivers and other drivers on the road. Due to this need…

Self-driving cars must be able to detect objects quickly and accurately to ensure the safety of their drivers and other drivers on the road. Due to this need for real-time processing in autonomous driving (AD) and visual inspection use cases, multiple AI models with preprocessing and postprocessing logic are combined in a pipeline and used for machine learning (ML) inference.

Speedup is required in every step of the pipeline to ensure a low latency workflow. Latency is the time it takes to get the inference response. Faster processing of AD data will enable more efficient analysis and use of the information, creating a safer driving environment. A delay with any single aspect can slow down the entire pipeline. 

To achieve a low latency inference workflow, electric vehicle manufacturer NIO integrated NVIDIA Triton Inference Server into their AD inference pipeline. NVIDIA Triton Inference Server is an open source multiframework inference serving software.

This post explains how NIO orchestrated its pipeline of image preprocessing and postprocessing and AI models with NVIDIA Triton on GPUs. It also shows how NIO reduced network transmission to successfully speed up their AI inference workflow for AD use cases.

Join the NVIDIA Triton and NVIDIA TensorRT community and stay current on the latest product updates, bug fixes, content, best practices, and more.

Faster AI inference for real-time response

NIO designs, develops, jointly manufactures, and sells premium smart electric vehicles, driving innovations in next-generation technologies in autonomous driving, digital technologies, electric powertrains, and batteries. NIO Autonomous Driving Development Platform (NADP) is an R&D platform dedicated to the core autonomous driving service of NIO.

NIO chose NVIDIA Triton Inference Server because of several key technical and operational reasons, including:

• NVIDIA Triton supports DAG-based orchestration of numerous models, along with preprocessing or postprocessing modules
• Cloud-native deployment of NVIDIA Triton enabled multi-GPU, multi-node scaling in a lightweight way
• High-quality documentation and learning resources helped ease migration to NVIDIA Triton
• NVIDIA Triton’s stability and robust functionality are necessary for AD use cases 

NIO’s AI inference workflow for autonomous driving

Hundreds of AI models are used to mine data from autonomous vehicles. In a use case like autonomous driving, the inference workflow consists of multiple AI models with preprocessing and postprocessing logic stitched together in a pipeline. 

NIO moved the preprocessing and postprocessing of the pipeline from the client side, which runs on CPUs, to NVIDIA Triton running on GPUs. The NVIDIA Triton’s business logic scripting (BLS) functionality was used to orchestrate the pipeline to run optimally for AD use. 

By moving the preprocessing from CPU to GPU and leveraging efficient pipeline orchestration, NIO achieved 6x latency reduction in some core pipelines, improving the overall throughput by up to 5x. 

Before and after workflow pipelines are shown in Figure 1.

Model pipeline orchestration benefits of NVIDIA Triton

This section examines each of the benefits NIO realized by integrating NVIDIA Triton.

GPU-accelerated preprocessing

Preprocessing tasks such as decoding, resizing, and transposing were accelerated on the GPU by NVIDIA Triton using nvJPEG and NVIDIA DALI. This significantly offloaded the computing workload from the client CPU and reduced preprocessing latency.

Upgrading models without the need for client application modification

By moving the preprocessing and postprocessing of the model to NVIDIA Triton, each time the model is upgraded, the client side does not require any modification. This essentially speeds up the rollout of the model, helping it reach production faster.

Using a single GPU node to reduce network data transfer overhead

A unified preprocessing enables multiple copies of the input to be shared with multiple backend recognition models. The process uses GPU shared memory on the server side, without data transfer overhead costs. 

Figure 2 shows the pipeline can connect up to nine models using the NVIDIA Triton business logic scripting functionality.

For an input image of 2 K resolution, the size of each frame is 1920 x 1080 x 3 x 8 = 47 Mb. Assuming a full frame rate of 60 fps, the amount of data input per second is 1920 x 1080 x 3 x 8 x 60 = 2847 Mb. In the previous workflow, each image is sent sequentially to the nine models over the network. Data transferred per second is 1920 x 1080 x 3 x 8 x 60 x 9 = 25 Gb = 3 GB.

In the new workflow, the nine models are orchestrated with the NVIDIA Triton business logic scripting. That means the models can access the image in the GPU shared memory and the images do not have to be sent over the network. Assuming a PCIe bandwidth of 160 Gb = 20 GB per second, theoretically the data generated per second can save 150 ms in data transfer if the data is transferred over PCIe.

Assuming an available bandwidth of 16 Gb = 2 GB per second, theoretically the data generated per second can save 1,500 ms in data transfer if the data is transferred over the network. All these result in speeding up the workflow.

Network transfer savings using image compression 

For accurate model prediction, the input image must be 1920 x 1080 x 3 x 8 bytes in the previous workflow and must be transmitted through the network. After introducing the server-side preprocessing, the original image can be altered to a compressed three-channel 720 pixel image (1280 x 720 x 3) within the allowed range of accuracy loss. 

As a result, it only takes a few hundred KB to transmit the bytes of the compressed image and resize with minimal accuracy loss to 1920 x 1080 x 3 x 8 bytes on the server. This leads to additional network transfer savings, speeding up the workflow. 

Ease of integration in NADP inference platform

NIO’s current inference platform based on NVIDIA Triton is a key component of their Autonomous Driving Development Platform (NADP), used in their autonomous driving solution.

As the NIO platform is built on Kubernetes (K8s), it was imperative for NVIDIA Triton to integrate well with Kubernetes. The components of the workflow are implemented as K8s CRD (native and custom) around NVIDIA Triton.

Continuous Integration/Continuous Delivery (CI/CD)

Argo is the engine used to orchestrate the workflow in Kubernetes. It helps with CI/CD for all the components involved in development, quantification, access, cloud deployment, pressure testing, and launch. NVIDIA Triton helps with CI/CD by triggering the next step in the workflow whenever the models are loaded.

In addition, use of the NVIDIA Triton Docker container helps with consistent functionality across development, test, and deployment environments.

Integrating the Jupyter environment into the NVIDIA Triton image was seamless. Jupyter provides a convenient development environment for debugging in case of a complex problem that requires online debugging or offline reproduction.

Ease of deployment with Istio

NVIDIA Triton natively supports gRPC protocol for communication with applications. However, as the Kubernetes native service cannot offer effective request-level load balancing for gRPC, NVIDIA Triton is integrated with the Istio service mesh. Istio is used to load balance traffic to NVIDIA Triton Inference Server and monitor the health of the service through liveness/readiness probes of NVIDIA Triton.

Ease of use with Apollo configuration management

Apollo Configuration Center is used for model name-based service discovery. Users can access the models without knowing the specific domain name where the model is deployed. Combined with the NVIDIA Triton model repository, users can directly trigger the deployment of models.

Metrics with Prometheus and Grafana

NVIDIA Triton provides a complete set of model service metrics based on model dimensions. For example, NVIDIA Triton can distinguish between inference request queueing time and GPU computation time, enabling fine-grained diagnosis and analysis of online model service performance without entering the debug mode. 

Because NVIDIA Triton supports cloud-native mainstream Prometheus/Grafana, users can easily configure the dashboard and the alarms for each dimension to provide metrics support for high service availability.

Key takeaways

NIO’s optimized workflow that integrates NVIDIA Triton Inference Server resulted in a 6x latency reduction in some core pipelines. This improved overall throughput by up to 5x.

By moving the preprocessing logic to GPU using the NVIDIA Triton pipeline orchestration functionality, NIO achieved:

• Faster image processing
• Freed CPU capacity
• Reduced network transfer overhead
• Higher inference throughput

NIO achieved AI inference workflow acceleration using NVIDIA Triton Inference Server. NVIDIA Triton was also easy to integrate in a robust Kubernetes-based scalable solution.

Additional resources

• Get started with NVIDIA Triton and access a variety of beginner to advanced resources.
• Learn about the features needed in an inference platform when building a real-time or continuous data streaming application with Fast and Scalable AI Model Deployment with NVIDIA Triton Inference Server.
• Watch the NVIDIA On-Demand session, From Cloud to Car: How NIO Develops Intelligent Vehicles.

Timing is everything, especially when it impacts your customer experiences, bottom line, and production efficiency. Edge AI can help by delivering real-time…

Timing is everything, especially when it impacts your customer experiences, bottom line, and production efficiency. Edge AI can help by delivering real-time intelligence and increased privacy in intermittent, low bandwidth, and low cost environments. 

By 2025, according to Gartner®, 75% of data will be created and processed at the edge, outside the traditional data center or cloud.¹ It’s no wonder that thousands of companies are turning to edge AI to drive transformation for their businesses.

As organizations undergo this shift, many IT and business leaders are still in the early stages of planning and executing their edge computing strategies. Because edge AI is a new concept, the process is difficult for many. 

NVIDIA, a leading AI infrastructure company with robust experience helping organizations, customers, and partners successfully deploy edge AI solutions, is no stranger to these new concepts. 

In an effort to help others, the learnings and recommendations from these experiences are presented in An IT Manager’s Guide: How to Successfully Deploy an Edge AI Solution. The whitepaper offers an in-depth look at building and executing a successful edge AI deployment.

This post features recommendations on some key considerations when configuring an edge system. 

Edge system configurations: Design recommendations 

There are many parameters to consider when sizing a system. The optimal PCIe server configuration will depend on the target workload for that server. 

Edge AI models incorporate various workloads into their applications, such as vision AI, natural language processing, recommendations based on industrial sensors, and predictive analytics. 

Edge computing sizing considerations

When it comes to designing full hardware and software solutions at the edge, it is important to look at the solution as a whole to understand how the parts work together. Below are some of the individual considerations that IT must evaluate for edge AI deployments.

Number of streams: Each camera feed is a stream requiring a certain amount of memory and compute for processing. Small configurations of 6-7 video processing streams require relatively small systems. Larger deployments may require high performance systems that are typically seen in the data center. 

Application examples: One of the first steps to a successful edge AI deployment is understanding what workload needs to be run to reach your goals. Vision AI applications like image recognition, people or vehicle detection, and segmentation are all common use cases. 

Once an application is determined, it is important to understand the intended scale. For example, are additional AI models needed? Typically, a proof of concept (POC) will consist of a single AI model and use case, but most production deployments ultimately incorporate multiple AI models. The next steps include quantifying the business value of the application, dictating any environmental constraints, and securing stakeholder alignment. 

Memory: Perhaps the most common way to under-resource an edge AI solution is to configure the edge systems with too little memory. Edge AI systems require significantly more memory than other applications to support the parallel execution of the inference engine across the CPU and GPUs. 

The data science team or application vendor who trained the AI will know the memory requirements of the latest model. IT teams should, at a minimum, double that number to accommodate the inevitable expansion of the model as it retrains. This will also provide some headroom for the additional AI models that will need to be deployed alongside the first one.

Another rule of thumb is to provision twice as much system memory as the total GPU memory, and never less than 1.5x the total GPU memory. The memory should be evenly spread across all CPU sockets and memory channels for optimal performance.

Networking: As operations increasingly rely on digital technologies such as edge computing, resilience is key. There are two networks to consider when designing an edge solution: the network between the edge AI location and cloud, and the network between a sensor and the edge AI system. 

Understanding the type of network connectivity of your environment will help in determining the specific networking bandwidth requirements for your use case. For example, for a use case like robotics, where wireless connectivity may not be possible, 5G is the next best choice as it offers minimal congestion and guaranteed service and bandwidth. 

Accelerators: Most edge applications run adequately on single socket x86 or Arm CPUs. But when the edge applications incorporate AI capabilities, they are far more compute intensive. 

To run an inference engine at the edge, the edge hardware needs enough compute power to execute complex neural networks with massively parallel computations. CPUs execute all the independent cells of a neural network sequentially, while discreet accelerators can execute them in parallel. Hence, accelerators are architecturally suited for AI, providing meaningfully better performance. They have become an essential component of modern AI infrastructure.

Among the most effective discrete accelerators for edge AI are Graphics Processing Units (GPUs) and Data Processing Units (DPUs). 

Storage: Naturally, the edge server requires local storage, usually a solid state hard drive, for its operating system, network components, hardware drivers, and application software. Unlike other applications, edge AI solutions typically process a massive amount of unstructured input data such as images, voices, and sensor readings. Depending upon how much of this data needs to be stored, for how long, how securely, and the level of reliability, different storage options are called for.

The first step in determining what storage is necessary for an edge AI solution requires IT teams to think through a data strategy. The data strategy will dictate what and how much data will need to be stored locally or in the cloud, and in turn, guide what storage options are best for that particular solution. Without a proactive strategy, developers often make inconsistent and sub-optimal choices that create problems down the road. 

Security: Security is paramount for edge AI computing devices, as they are deployed in remote locations outside the data center firewalls and the physical protections that limit access to systems. For more details, see Edge Computing: Considerations for Security Architects.

When it comes to an edge AI solution, five areas should be understood and made part of the overall solutions architecture: end-to-end encryption, mutual authentication, physical security, zero trust networking, and real-time monitoring.

Management: A remote management plan is critical for edge environments because systems at the edge are distributed, always on, and often operate in remote settings. See Remotely Operating Systems and Applications at the Edge to learn more.

An edge management solution will have automatic deployment and provisioning capabilities, ongoing management, real-time alerting, auditing, and use modern, cloud-native tools. 

Organizations have the choice of whether to build or buy a management solution. The following are questions to consider: How quickly does a solution need to be set up? Is the appropriate team and expertise available? Does this provide secure management of my edge environment? 

Pillars of a successful edge deployment 

Deploying the infrastructure needed to support a scalable edge AI solution is a big challenge. The process is iterative and time consuming, yet critical to do correctly. Decisions that are made when building an edge AI solution have far-reaching implications that will impact an organization’s business outcomes. 

For more guidance on this topic, download An IT Manager’s Guide: How to Successfully Deploy an Edge AI Solution. 

References

1. Gartner, “Building an Edge Computing Strategy,” G00753920, September 2021. GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. All rights reserved.