1604164740
Here’s Why You Must Attend The DLDC 2020 — The Deep Learning Conference Of The Year
Organised by the Association of Data Scientists (ADaSci), DLDC 2020, or also known as the Deep Learning DevCon 2020, is another conference of the year that is hosted in partnership with Analytics India Magazine. Scheduled for 29th and 30th October, DLDC will bring together the leading experts and best minds of deep learning and machine learning industry from around the world. This two-day conference will witness an exciting lineup of extraordinary speakers, interesting talks and practical hands-on that will be interesting for professionals of this field as well for the students and enthusiasts who are looking to kickstart their deep learning career.
Organised by a premier global professional body of data science and machine learning professionals — ADaSci, this event will be the first-of-its-kind virtual conference on deep learning. With deep learning becoming one of the most advancing technologies in the world — from being used in the fields of natural language processing to making self-driving cars — this deep learning conference will aim at bridging the gap between the latest technological advancements in the research area and real-world applications of the same.
If you are ready to understand how to apply deep learning to your businesses, this conference is a must-attend. In this article, we will share a few more reasons as to why you should attend DLDC 2020.
#machine-learning
What is GEEK
Buddha Community
1603735200
Top 10 Deep Learning Sessions To Look Forward To At DVDC 2020
The Deep Learning DevCon 2020, DLDC 2020, has exciting talks and sessions around the latest developments in the field of deep learning, that will not only be interesting for professionals of this field but also for the enthusiasts who are willing to make a career in the field of deep learning. The two-day conference scheduled for 29th and 30th October will host paper presentations, tech talks, workshops that will uncover some interesting developments as well as the latest research and advancement of this area. Further to this, with deep learning gaining massive traction, this conference will highlight some fascinating use cases across the world.
Here are ten interesting talks and sessions of DLDC 2020 that one should definitely attend:
Also Read: Why Deep Learning DevCon Comes At The Right Time
Adversarial Robustness in Deep Learning
By Dipanjan Sarkar
**About: **Adversarial Robustness in Deep Learning is a session presented by Dipanjan Sarkar, a Data Science Lead at Applied Materials, as well as a Google Developer Expert in Machine Learning. In this session, he will focus on the adversarial robustness in the field of deep learning, where he talks about its importance, different types of adversarial attacks, and will showcase some ways to train the neural networks with adversarial realisation. Considering abstract deep learning has brought us tremendous achievements in the fields of computer vision and natural language processing, this talk will be really interesting for people working in this area. With this session, the attendees will have a comprehensive understanding of adversarial perturbations in the field of deep learning and ways to deal with them with common recipes.
Read an interview with Dipanjan Sarkar.
Imbalance Handling with Combination of Deep Variational Autoencoder and NEATER
By Divye Singh
**About: **Imbalance Handling with Combination of Deep Variational Autoencoder and NEATER is a paper presentation by Divye Singh, who has a masters in technology degree in Mathematical Modeling and Simulation and has the interest to research in the field of artificial intelligence, learning-based systems, machine learning, etc. In this paper presentation, he will talk about the common problem of class imbalance in medical diagnosis and anomaly detection, and how the problem can be solved with a deep learning framework. The talk focuses on the paper, where he has proposed a synergistic over-sampling method generating informative synthetic minority class data by filtering the noise from the over-sampled examples. Further, he will also showcase the experimental results on several real-life imbalanced datasets to prove the effectiveness of the proposed method for binary classification problems.
Default Rate Prediction Models for Self-Employment in Korea using Ridge, Random Forest & Deep Neural Network
By Dongsuk Hong
About: This is a paper presentation given by Dongsuk Hong, who is a PhD in Computer Science, and works in the big data centre of Korea Credit Information Services. This talk will introduce the attendees with machine learning and deep learning models for predicting self-employment default rates using credit information. He will talk about the study, where the DNN model is implemented for two purposes — a sub-model for the selection of credit information variables; and works for cascading to the final model that predicts default rates. Hong’s main research area is data analysis of credit information, where she is particularly interested in evaluating the performance of prediction models based on machine learning and deep learning. This talk will be interesting for the deep learning practitioners who are willing to make a career in this field.
#opinions #attend dldc 2020 #deep learning #deep learning sessions #deep learning talks #dldc 2020 #top deep learning sessions at dldc 2020 #top deep learning talks at dldc 2020
1671543249
Migrating From Jira Server: Guide, Pros, And Cons
February 15, 2022 marked a significant milestone in Atlassian’s Server EOL (End Of Life) roadmap. This was not the final step. We still have two major milestones ahead of us: end of new app sales in Feb 2023, and end of support in Feb 2024. In simpler words, businesses still have enough time to migrate their Jira Server to one of the two available products – Atlassian Cloud or Atlassian DC. But the clock is ticking.
Jira Cloud VS Data Center
If we were to go by Atlassian numbers, 95% of their new customers choose cloud.
“About 80% of Fortune 500 companies have an Atlassian Cloud license. More than 90% of new customers choose cloud first.” – Daniel Scott, Product Marketing Director, Tempo
So that’s settled, right? We are migrating from Server to Cloud? And what about the solution fewer people talk about yet many users rely on – Jira DC?
Both are viable options and your choice will depend greatly on the needs of your business, your available resources, and operational processes.
Let’s start by taking a look at the functionality offered by Atlassian Cloud and Atlassian DC.
| Feature | Atlassian Cloud | Atlassian Data Center |
| Product Plans | Multiple plans | One plan |
| Billing | Monthly and annual | Annual only |
| Pricing model | Per user or tiered | Tiered only |
| Support | Varying support levels depending on your plan: Enterprise support coverage is equivalent to Atlassian’s Data Center Premier Support offering | Varying support levels depending on the package: Priority Support or Premier Support (purchased separately) |
| Total Cost of Ownership | TCO includes your subscription fee, plus product administration time | TCO includes your subscription fee and product administration time, plus: costs related to infrastructure provisioning or IaaS fees (for example, AWS costs) planned downtime time and resources needed for software upgrades |
| Data encryption services |  |  |
| Data residency services | | |
| Audit logging | Organization-level audit logging available via Atlassian Access (Jira Software, Confluence) Product-level audit logs (Jira Software, Confluence) | Advanced audit logging |
| Device security | Mobile device management support (Jira Software, Confluence, Jira Service Management) Mobile application management (currently on the roadmap) | Mobile device management support (Jira Software, Confluence, Jira Service Management) |
| Content security | | |
| Data Storage limits | 2 GB (Free) 250 GB (Standard) Unlimited storage (Premium and Enterprise) | No limits |
| Performance | Continuous performance updates to improve load times, search responsiveness, and attachments Cloud infrastructure hosted in six geographic regions to reduce latency | Rate limitingCDN supports Smart mirrors and mirror farms (Bitbucket) |
| Backup and data disaster recovery | Jira leverages multiple geographically diverse data centers, has a comprehensive backup program, and gains assurance by regularly testing their disaster recovery and business continuity plans. Backups are generated daily and retained for 30 days to allow for point-in-time data restoration | |
| Containerization and orchestration | | Docker images Kubernetes support (on the roadmap for now) |
| Change management and upgrades | Atlassian automatically handles software and security upgrades for you Sandbox instance to test changes (Premium and Enterprise) Release track options for Premium and Enterprise (Jira Software, Jira Service Management, Confluence) | |
| Direct access to the database | No direct access to change the database structure, file system, or other server infrastructure Extensive REST APIs for programmatic data access | Direct database access |
| Insights and reporting | Organization and admin insights to track adoption of Atlassian products, and evaluate the security of your organization. | Data Pipeline for advanced insightsConfluence analytics |
Pros and cons of Jira Cloud
When talking about pros and cons, there’s always a chance that a competitive advantage for some is a dealbreaker for others. That’s why I decided to talk about pros and cons in matching pairs.
Pro: Scalability is one of the primary reasons businesses are choosing Jira Cloud. DC is technically also scalable, but you’ll need to scale on your own whereas the cloud version allows for the infrastructure to scale with your business.
Con: Despite the cloud’s ability to grow with your business, there is still a user limit of 35k users. In addition to that, the costs will grow alongside your needs. New users, licenses, storage, and computing power – all come at an additional cost. So, when your organization reaches a certain size, migrating to Jira DC becomes more cost-efficient.
Pro: Jira takes care of maintenance and support for you.
Con: Your business can suffer from unpredicted downtime. And there are certain security risks.
Pro: Extra bells and whistles:
- Sandbox: Sandbox is a safe environment system admins can use to test applications and integrations before rolling them out to the production environment.
- Release tracks: Admins can be more flexible with their product releases as they can access batch and control cloud releases. This means they’ll have much more time to test existing configurations and workflows against a new update.
- Insight Discovery: More data means more ways you can impact your business or product in a positive, meaningful way.
- Team Calendars: This is a handy feature for synchronization and synergy across teams.
Con: Most of these features are locked behind a paywall and are only available to either Premium and Enterprise or only Enterprise licenses (either fully or through addition of functionality. For example, Release tracks are only available to Enterprise customers.) In addition, the costs will grow as you scale the offering to fit your growing needs.
Pros and cons of Jira Data Center
I’ll be taking the same approach to talking about the pros and cons as I did when writing about Atlassian Cloud. Pros and cons are paired.
Pro: Hosting your own system means you can scale horizontally and vertically through additional hardware. Extension of your systems is seamless, and there is no downtime (if you do everything correctly). Lastly, you don’t have to worry about the user limit – there is none.
Con: While having more control over your systems is great, it implies a dedicated staff of engineers, additional expenses on software licensing, hardware, and physical space. Moreover, seamless extension and 0% downtime are entirely on you.
Pro: Atlassian has updated the DC offering with native bundled applications such as Advanced Roadmaps, team calendars and analytics for confluence, insight asset management, and insight discovery in Jira Service Management DC.
Con: Atlassian has updated their pricing to reflect these changes. And you are still getting fewer “bells and whistles” than Jira Cloud users (as we can see from the feature comparison).
Pro: You are technically safer as the system is supported on your hardware by your specialists. Any and all Jira server issues, poor updates, and downtime are simply not your concern.
Con: Atlassian offers excellent security options: data encryption in transit and rest, to mobile app management, to audit offerings and API token controls. In their absence, your team company has to dedicate additional resources to security.
Pro: Additional benefits from Atlassian, such as the Priority Support bundle (all DC subscriptions have this option), and the Data center loyalty discount (more on that in the pricing section.)
The Pricing
Talking about pricing of SaaS products is always a challenge as there are always multiple tiers and various pay-as-you go features. Barebones Jira Cloud, for instance, is completely free of charge, yet there are a series of serious limitations.
Standard Jira Cloud will cost you an average of $7.50 per user per month while premium cranks that price up to $14.50. The Enterprise plan is billed annually and the cost is determined on a case-by-case basis. You can see the full comparison of Jira Cloud plans here. And you can use this online calculator to learn the cost of ownership in your particular case.
| 50 Users | Standard (Monthly/Annually) | Premium (Monthly/Annually) |
| Jira Software | $387.50 / $3,900 | $762.50 / $7,650 |
| Jira Work Management | $250 / $2,500 | |
| Jira Service Management | $866.25 / $8,650 | $2,138.25 / $21,500 |
| Confluence | $287.50 / $2,900 | $550 / $5,500 |
| 100 Users | Standard (Monthly/Annually) | Premium (Monthly/Annually) |
| Jira Software | $775 / $7,750 | $1,525 / $15,250 |
| Jira Work Management | $500 / $5,000 | |
| Jira Service Management | $1,653.75 / $16,550 | $4,185.75 / $42,000 |
| Confluence | $575 / $5,750 | $1,100 / $11,000 |
| 500 Users | Standard (Monthly/Annually) | Premium (Monthly/Annually) |
| Jira Software | $3,140 / $31,500 | $5,107.50 / $51,000 |
| Jira Work Management | $1,850 / $18,500 | |
| Jira Service Management | $4,541.25 / $45,400 | $11,693.25 / $117,000 |
| Confluence | $2,060 / $20,500 | $3,780 / $37,800 |
Please note that these prices were calculated without any apps included.
Jira Data Center starts at $42,000 per year and the plan includes up to 500 users. If you are a new client and are not eligible for any discounts*, here’s a chart that should give you an idea as to the cost of ownership of Jira DC. You can find more information regarding your specific case here.
| Users | Commercial Annual Plan | Academic Annual Plan |
| 1-500 | USD 42,000 | USD 21,000 |
| 501-1000 | USD 72,000 | USD 36,000 |
| 1001-2000 | USD 120,000 | USD 60,000 |
| Confluence for Data Center | ||
| 1-500 | USD 27,000 | USD 13,500 |
| 501-1000 | USD 48,000 | USD 24,000 |
| 1001-2000 | USD 84,000 | USD 42,000 |
| Bitbucket for Data Center | ||
| 1-25 | USD 2,300 | USD 1,150 |
| 26-50 | USD 4,200 | USD 2,100 |
| 51-100 | USD 7,600 | USD 3,800 |
| Jira Service Management for Data Center | ||
| 1-50 | USD 17,200 | USD 8,600 |
| 51-100 | USD 28,600 | USD 14,300 |
| 101-250 | USD 51,500 | USD 25,750 |
*Discounts:
- Centralized per-user licensing allows users access all enterprise instances with a single Enterprise license.
- There’s an option for dual licensing for users who purchase an annual cloud subscription with 1,001 or more users. In this case, Atlassian extends your existing server maintenance or Data Center subscription for up to one year at a 100% discount.
- There are certain discounts for apps depending on your partnership level.
- Depending on your situation, you may qualify for several Jira Data Center discount programs:
- Academic Data Center licensing offers a 50% discount.
- Community Data Center licensing for Non-profit organizations is free of charge.
- Lastly, there is a loyalty discount for Server customers who choose to migrate to the Data Center. Check this link to find out if you are eligible.
What should be your User Migration strategy?
Originally, there were several migration methods: Jira Cloud Migration Assistant, Jira Cloud Site Import, and there was an option to migrate via CSV export (though Jira actively discourages you from using this method). However, Jira’s team has focused their efforts on improving the Migration Assistant and have chosen to discontinue Cloud Site Import support.
Thanks to the broadened functionality of the assistant, it is now the only go-to method for migration with just one exception. If you are migrating over 1000 users and you absolutely need to migrate advanced roadmaps – you’ll need to rely on Site Import. At least for now, as Jira is actively working on implementing this feature in their assistant.
Here’s a quick comparison of the options and their limitations.
| Features | Limitations | |
| Cloud Migration Assistant | App migration Existing data on a Cloud Site is not overwritten You choose the projects, users, and groups you want to migrate Jira Service Management customer account migration Better UI to guide you through the migration Potential migration errors are displayed in advance Migration can be done in phases reducing the downtime Pre- and post-migration reports | You must be on a supported self-managed version of Jira |
| Site Export | Can migrate Advanced Roadmaps | App data is not migrated Migration overrides existing data on the Cloud site Separate user importUsers from external directories are not migrated No choice of data you want or don’t want migrated There’s a need to split attachments into up to 5GB chunks Higher risks of downtime due to the “all or nothing” approach You must be on a supported self-managed version of Jira |
Pro tip: If you have a large base of users (above 2000), migrate them before you migrate projects and spaces. This way, you will not disrupt the workflow as users are still working on Server and the latter migration of data will take less time.
How to migrate to Jira Cloud
Now that we have settled on one particular offering based on available pricing models as well as the pros and the cons that matter the most to your organization, let’s talk about the “how”.
How does one migrate from Jira Server to Jira Cloud?
Pre-migration checklist
Jira’s Cloud Migration Assistant is a handy tool. It will automatically review your data for common errors. But it is incapable of doing all of the work for you. That’s why we – and Atlassian for that matter – recommend creating a pre-migration checklist.
Smart Checklist will help you craft an actionable, context-rich checklist directly inside a Jira ticket. This way, none of the tasks will be missed, lost, or abandoned.
Below is an example of how your migration checklist will look like in Jira.
Feel free to copy the code and paste it into your Smart Checklist editor and you’ll have the checklist at the ready.
# Create a user migration plan #must
> Please keep in mind that Jira Cloud Migration Assistant migrates all users and groups as well as users and groups related to selected projects
- Sync your user base
- Verify synchronization
- External users sync verification
- Active external directory verification
## Check your Jira Server version #must
- Verify via user interface or Support Zip Product Version Verification
> Jira Migration Assistant will not work unless Jira is running on a supported version
## Fix any duplicate email addresses #must
- Verify using SQL
> Duplicate email addresses are not supported by Jira Cloud and therefore can't be migrated with the Jira Cloud Migration Assistant. To avoid errors, you should find and fix any duplicate email addresses before migration. If user information is managed in an LDAP Server, you will need to update emails there and sync with Jira before the migration. If user information is managed locally, you can fix them through the Jira Server or Data Center user interface.
## Make sure you have the necessary permissions #must
- System Admin global permissions on the Server instance
- Exists in the target Cloud site
- Site Administrator Permission in the cloud
## Check for conflicts with group names #must
- Make sure that the groups in your Cloud Site don't have the same names as groups in Server
> Unless you are actively trying to merge them
- Delete or update add-on users so not to cause migration issues
- Verify via SQL
## Update firewall allowance rules #must
- None of the domains should be blocked by firewall or proxy
## Find a way to migrate apps #must
- Contact app vendors
## Check public access settings #must
- Projects
- Filters
- Filters
- Boards
- Dashboards
## Review server setup #mst
- at least 4gb Heap Allocation
- Open Files limit review
- Verify via support zip
## Check Server timezone #must for merging Cloud sites
- Switch to UTC is using any other timezone
> Add a system flag to the Jira Server instance -Duser.timezone=UTC as outlined in this article about updating documentation to include timezone details.
## Fix any duplicate shared configuration
## Storage limits
## Prepare the server instance
- Check data status
- All fields have value and are not null
-Any archived projects you wish to migrate are activated
## Prepare your cloud site
- Same Jira products enabled
- Same language
- User migration strategy
## Data backup
- Backup Jira Server site
- Backup Cloud site
## Run a test migration
- Done
## Notify Jira support
- Get in touch with Jira migration supportUse backups
On the one hand, having all of your Jira products on a server may seem like a backup in and of itself. On the other hand, there are data migration best practices we should follow even if it’s just a precaution. No one has ever felt sorry for their data being too safe.
In addition, there are certain types of migration errors that can be resolved much faster with having a backup at hand.
- Jira Server Database backup: this step creates a DB backup in an XML format.
- Log in with Jira System Admin permissions
- Go to system -> Import and Export -> Backup Manager -> Backup for server.
- Click the create Backup for server button.
- Type in the name for your backup.
- Jira will create a zipped XML file and notify you once the backup is ready.
- Jira Cloud Backup: This backup also saves your data in an XML format. The process is quite similar to creating a Jira Server backup with the only difference taking place on the Backups page.
- Select the option to save your attachments, logos, and avatars.
- Click on the Create backup button.
- As you can see, the Cloud backup includes the option to save attachments, avatars, and logos. This step should be done manually when backing up Server data.
- Create a Zip archive for this data
- Make sure it follows the structure suggested by Atlassian
Migrating your Jira instance to the cloud via the Jira Migration Assistant
Jira Cloud Migration Assistant is a free add-on Atlassian recommends using when migrating to the cloud. It accesses and evaluates your apps and helps migrate multiple projects.
Overall, the migration assistant offers a more stable and reliable migration experience. It automatically checks for certain errors. It makes sure all users have unique and valid emails, and makes sure that none of the project names and keys conflict with one another.
This is a step-by-step guide for importing your Jira Server data backup file into Jira Cloud.
- Log into Jira Cloud with admin permissions
- Go to System -> Import and Export -> External System Import
- Click on the Jira Server import option
- Select the backup Zip you have created
- Jira will check the file for errors and present you with two options: enable or disable outgoing mail. Don’t worry, you will be able to change this section after the migration process is complete.
- Then you will be presented with an option to merge Jira Server and Jira Cloud users
- Choosing overwrite will replace the users with users from the imported files
- The merge option will merge groups with the same name
- Lastly, you can select the third option if you are migrating users via Jira’s assistant
- Run the import
How do you migrate Jira Server into Jira DC?
Before we can proceed with the migration process, please make sure you meet the following prerequisites:
- Make sure you are installing Jira on one of the supported platforms. Atlassian has a list of supported platforms for Jira 9.1.
- Make sure the applications you are using are compatible with Jira DC. You will be required to switch to datacenter-compatible versions of your applications (they must be available).
- Make sure you meet the necessary software and hardware requirements:
- You have a DC license
- You are using a supported database, OS, and Java version
- You are using OAuth authentication if your application links to other Atlassian products
Once you are certain you are ready to migrate your Jira Server to Jira Data Center, you can proceed with an installation that’s much simpler than one would expect.
- Upgrade your apps to be compatible with Jira DC
- Go to Administration -> Applications -> Versions and licenses
- Enter your Jira DC License Key
- Restart Jira
That’s it. You are all set. Well, unless your organization has specific needs such as continuous uptime, performance under heavy loads, and scalability, in which case you will need to set up a server cluster. You can find out more about setting up server clusters in this guide.
1618317562
Top Deep Learning Development Services | Hire Deep Learning Developer
View more: https://www.inexture.com/services/deep-learning-development/
We at Inexture, strategically work on every project we are associated with. We propose a robust set of AI, ML, and DL consulting services. Our virtuoso team of data scientists and developers meticulously work on every project and add a personalized touch to it. Because we keep our clientele aware of everything being done associated with their project so there’s a sense of transparency being maintained. Leverage our services for your next AI project for end-to-end optimum services.
#deep learning development #deep learning framework #deep learning expert #deep learning ai #deep learning services
1594753020
Citrix Bugs Allow Unauthenticated Code Injection, Data Theft
Multiple vulnerabilities in the Citrix Application Delivery Controller (ADC) and Gateway would allow code injection, information disclosure and denial of service, the networking vendor announced Tuesday. Four of the bugs are exploitable by an unauthenticated, remote attacker.
The Citrix products (formerly known as NetScaler ADC and Gateway) are used for application-aware traffic management and secure remote access, respectively, and are installed in at least 80,000 companies in 158 countries, according to a December assessment from Positive Technologies.
Other flaws announced Tuesday also affect Citrix SD-WAN WANOP appliances, models 4000-WO, 4100-WO, 5000-WO and 5100-WO.
Attacks on the management interface of the products could result in system compromise by an unauthenticated user on the management network; or system compromise through cross-site scripting (XSS). Attackers could also create a download link for the device which, if downloaded and then executed by an unauthenticated user on the management network, could result in the compromise of a local computer.
“Customers who have configured their systems in accordance with Citrix recommendations [i.e., to have this interface separated from the network and protected by a firewall] have significantly reduced their risk from attacks to the management interface,” according to the vendor.
Threat actors could also mount attacks on Virtual IPs (VIPs). VIPs, among other things, are used to provide users with a unique IP address for communicating with network resources for applications that do not allow multiple connections or users from the same IP address.
The VIP attacks include denial of service against either the Gateway or Authentication virtual servers by an unauthenticated user; or remote port scanning of the internal network by an authenticated Citrix Gateway user.
“Attackers can only discern whether a TLS connection is possible with the port and cannot communicate further with the end devices,” according to the critical Citrix advisory. “Customers who have not enabled either the Gateway or Authentication virtual servers are not at risk from attacks that are applicable to those servers. Other virtual servers e.g. load balancing and content switching virtual servers are not affected by these issues.”
A final vulnerability has been found in Citrix Gateway Plug-in for Linux that would allow a local logged-on user of a Linux system with that plug-in installed to elevate their privileges to an administrator account on that computer, the company said.
#vulnerabilities #adc #citrix #code injection #critical advisory #cve-2020-8187 #cve-2020-8190 #cve-2020-8191 #cve-2020-8193 #cve-2020-8194 #cve-2020-8195 #cve-2020-8196 #cve-2020-8197 #cve-2020-8198 #cve-2020-8199 #denial of service #gateway #information disclosure #patches #security advisory #security bugs
1593529260
Few Shot Learning — A Case Study (2)
In the previous blog, we looked into the fact why Few Shot Learning is essential and what are the applications of it. In this article, I will be explaining the Relation Network for Few-Shot Classification (especially for image classification) in the simplest way possible. Moreover, I will be analyzing the Relation Network in terms of:
- Effectiveness of different architectures such as Residual and Inception Networks
- Effects of transfer learning via using pre-trained classifier on ImageNet dataset
Moreover, effectiveness will be evaluated on the accuracy, time required for training, and the number of required training parameters.
Please watch the GitHub repository to check out the implementations and keep updated with further experiments.
Introduction to Few-Shot Classification
In few shot classification, our objective is to design a method which can identify any object images by analyzing few sample images of the same class. Let’s the take one example to understand this. Suppose Bob has a client project to design a 5 class classifier, where 5 classes can be anything and these 5 classes can even change with time. As discussed in previous blog, collecting the huge amount of data is very tedious task. Hence, in such cases, Bob will rely upon few shot classification methods where his client can give few set of example images for each classes and after that his system can perform classification young these examples with or without the need of additional training.
In general, in few shot classification four terminologies (N way, K shot, support set, and query set) are used.
- N way: It means that there will be total N classes which we will be using for training/testing, like 5 classes in above example.
- K shot: Here, K means we have only K example images available for each classes during training/testing.
- Support set: It represents a collection of all available K examples images from each classes. Therefore, in support set we have total N*K images.
- Query set: This set will have all the images for which we want to predict the respective classes.
At this point, someone new to this concept will have doubt regarding the need of support and query set. So, let’s understand it intuitively. Whenever humans sees any object for the first time, we get the rough idea about that object. Now, in future if we see the same object second time then we will compare it with the image stored in memory from the when we see it for the first time. This applied to all of our surroundings things whether we see, read, or hear. Similarly, to recognise new images from query set, we will provide our model a set of examples i.e., support set to compare.
And this is the basic concept behind Relation Network as well. In next sections, I will be giving the rough idea behind Relation Network and I will be performing different experiments on 102-flower dataset.
About Relation Network
The Core idea behind Relation Network is to learn the generalized image representations for each classes using support set such that we can compare lower dimensional representation of query images with each of the class representations. And based on this comparison decide the class of each query images. Relation Network has two modules which allows us to perform above two tasks:
- Embedding module: This module will extract the required underlying representations from each input images irrespective of the their classes.
- Relation Module: This module will score the relation of embedding of query image with each class embedding.
Training/Testing procedure:
We can define the whole procedure in just 5 steps.
- Use the support set and get underlying representations of each images using embedding module.
- Take the average of between each class images and get the single underlying representation for each class.
- Then get the embedding for each query images and concatenate them with each class’ embedding.
- Use the relation module to get the scores. And class with highest score will be the label of respective query image.
- [Only during training] Use MSE loss functions to train both (embedding + relation) modules.
Few things to know during the training is that we will use only images from the set of selective class, and during the testing, we will be using images from unseen classes. For example, from the 102-flower dataset, we will use 50% classes for training, and rest will be used for validation and testing. Moreover, in each episode, we will randomly select 5 classes to create the support and query set and follow the above 5 steps.
That is all need to know about the implementation point of view. Although the whole process is simple and easy to understand, I’ll recommend reading the published research paper, Learning to Compare: Relation Network for Few-Shot Learning, for better understanding.
#deep-learning #few-shot-learning #computer-vision #machine-learning #deep learning #deep learning