Freelancer logo
How It Works
Browse Jobs
Log In
Sign Up
Post a Project
You're now following .
Error following user.
This user does not allow users to follow them.
You are already following this user.
Your membership plan only allows 0 follows. Upgrade
here.
Successfully unfollowed user.
Error unfollowing user.
You have successfully recommended
Error recommending user.
Email successfully verified.
User Avatar
$30 USD / hour
Flag of INFlag of IN
ghasipura,
india
$30 USD / hour
It's currently 12:27 PM here
Joined July 28, 2020
0 Recommendations

Kaibalya B.

@jarviskb

annual-level-two.svg
4.9 (24 reviews)
4.9 (24 reviews)
5.2
5.2
$30 USD / hour
Flag of INFlag of IN
ghasipura,
india
$30 USD / hour
92%
Jobs Completed
87%
On Budget
33%
On Time
15%
Repeat Hire Rate

AI Researcher

Deep learning is my passion. I have 2 years experience in the field of machine learning and deep learning. Deep image processing and Natural language processing are my major field of expertise
Freelancer
Python Developers
India

Contact Kaibalya B. about your job

Log in to discuss any details over chat.

Portfolio Items

this model  does  better  than  its  competitors RoBERT and XLM-R and outperforms previous  SOTA  models  on  three  downstream Tweet NLP tasks of POS tagging, NER and text classification, thus confirming the effectiveness of the large-scale and domain-specific language model pre-trained for English Tweets. 

I am also attaching performance of this model on two benchmark dataset. POS   tagging   accuracy   results   on   the Ritter11-T-POS  (Ritter11),  ARK-Twitter  (ARK)  and TWEEBANK-V2  (TB-v2)  test  sets. And SemEval2018-Task3A  test  set.    F1pos—the  main  ranking  metric—denotes the F1 score computed for the positive label.
BERT model for English Tweets
this model  does  better  than  its  competitors RoBERT and XLM-R and outperforms previous  SOTA  models  on  three  downstream Tweet NLP tasks of POS tagging, NER and text classification, thus confirming the effectiveness of the large-scale and domain-specific language model pre-trained for English Tweets. 

I am also attaching performance of this model on two benchmark dataset. POS   tagging   accuracy   results   on   the Ritter11-T-POS  (Ritter11),  ARK-Twitter  (ARK)  and TWEEBANK-V2  (TB-v2)  test  sets. And SemEval2018-Task3A  test  set.    F1pos—the  main  ranking  metric—denotes the F1 score computed for the positive label.
BERT model for English Tweets
As we know the power of generative models in recent development. Starting from basic face creation from DCGAN to cyclic GAn and now style transfer with GAN these models have already proved their dominance in the field of generating fake images. Face swapping is an eye catching task in generating fake faces by transferring a source face to the destination faces while maintaining the facial features like moments, expressions and other information of the source. 
	
The key motivation behind all this task is generative models. More and more faces synthesized by StyleGAN, StyleGAN2 are becoming more and more realistic and completely indistinguishable to the human vision system.
Deep Fake
The architecture or dialogue management flow used in this research approach is as shown below which is the architecture of the chatbot stack framework itself. The first stage of the message is received and forwarded to the interpreter, namely Rasa NLU to extract intents,  entities, and other structured information. Both interpreter or tracker is tracking, detecting, and maintaining the status of the conversation context through the message notifications it has received.The third policy or policy manager receives context status from the tracker. The fourth policy or policy makers choose which action will be taken next. The five actions or actions are recorded by the tracker. These six actions are executed by sending a message to the user. Seventh,if the action that has been executed is ignored or ignored by the user at a certain time, the process returns to the third step.
Dialogue Management
The Rasa Stack is a set of open-source NLP tools focused primarily on chatbots. In fact, it’s one of the most effective and time efficient tools to build complex chatbots in minutes. Rasa Stack lets you focus on improving the “Chatbot” part of your project by providing readymade code for other background tasks like deploying, creating servers, etc. The default set up of Rasa works really well right out of the box for intent extraction and dialogue management, even with lesser data. Rasa stack is open-source, which means we know exactly what is happening under the hood and can customize things as much as we want. These features differentiate Rasa from other chatbot building platforms.
RASA NLU
High-resolution representations are essential for position-sensitive vision problems, such as semantic segmentation, and object detection. Existing state-of-the-art frameworks first encode the input image as a low-resolution representation through a subnetwork that is formed by connecting high-to-low resolution convolutions \emph{in series} (e.g., ResNet, VGGNet), and then recover the high-resolution representation from the encoded low-resolution representation. Instead, my proposed network, named as High-Resolution Network (HRNet), maintains high-resolution representations through the whole process. There are two key characteristics: (i) Connect the high-to-low resolution convolution streams \emph{in parallel}; (ii) Repeatedly exchange the information across resolutions. The benefit is that the resulting representation is semantically richer and spatially more precise.
Interior Segmentation
High-resolution representations are essential for position-sensitive vision problems, such as semantic segmentation, and object detection. Existing state-of-the-art frameworks first encode the input image as a low-resolution representation through a subnetwork that is formed by connecting high-to-low resolution convolutions \emph{in series} (e.g., ResNet, VGGNet), and then recover the high-resolution representation from the encoded low-resolution representation. Instead, my proposed network, named as High-Resolution Network (HRNet), maintains high-resolution representations through the whole process. There are two key characteristics: (i) Connect the high-to-low resolution convolution streams \emph{in parallel}; (ii) Repeatedly exchange the information across resolutions. The benefit is that the resulting representation is semantically richer and spatially more precise.
Interior Segmentation
High-resolution representations are essential for position-sensitive vision problems, such as semantic segmentation, and object detection. Existing state-of-the-art frameworks first encode the input image as a low-resolution representation through a subnetwork that is formed by connecting high-to-low resolution convolutions \emph{in series} (e.g., ResNet, VGGNet), and then recover the high-resolution representation from the encoded low-resolution representation. Instead, my proposed network, named as High-Resolution Network (HRNet), maintains high-resolution representations through the whole process. There are two key characteristics: (i) Connect the high-to-low resolution convolution streams \emph{in parallel}; (ii) Repeatedly exchange the information across resolutions. The benefit is that the resulting representation is semantically richer and spatially more precise.
Interior Segmentation
High-resolution representations are essential for position-sensitive vision problems, such as semantic segmentation, and object detection. Existing state-of-the-art frameworks first encode the input image as a low-resolution representation through a subnetwork that is formed by connecting high-to-low resolution convolutions \emph{in series} (e.g., ResNet, VGGNet), and then recover the high-resolution representation from the encoded low-resolution representation. Instead, my proposed network, named as High-Resolution Network (HRNet), maintains high-resolution representations through the whole process. There are two key characteristics: (i) Connect the high-to-low resolution convolution streams \emph{in parallel}; (ii) Repeatedly exchange the information across resolutions. The benefit is that the resulting representation is semantically richer and spatially more precise.
Interior Segmentation
High-resolution representations are essential for position-sensitive vision problems, such as semantic segmentation, and object detection. Existing state-of-the-art frameworks first encode the input image as a low-resolution representation through a subnetwork that is formed by connecting high-to-low resolution convolutions \emph{in series} (e.g., ResNet, VGGNet), and then recover the high-resolution representation from the encoded low-resolution representation. Instead, my proposed network, named as High-Resolution Network (HRNet), maintains high-resolution representations through the whole process. There are two key characteristics: (i) Connect the high-to-low resolution convolution streams \emph{in parallel}; (ii) Repeatedly exchange the information across resolutions. The benefit is that the resulting representation is semantically richer and spatially more precise.
Interior Segmentation
High-resolution representations are essential for position-sensitive vision problems, such as semantic segmentation, and object detection. Existing state-of-the-art frameworks first encode the input image as a low-resolution representation through a subnetwork that is formed by connecting high-to-low resolution convolutions \emph{in series} (e.g., ResNet, VGGNet), and then recover the high-resolution representation from the encoded low-resolution representation. Instead, my proposed network, named as High-Resolution Network (HRNet), maintains high-resolution representations through the whole process. There are two key characteristics: (i) Connect the high-to-low resolution convolution streams \emph{in parallel}; (ii) Repeatedly exchange the information across resolutions. The benefit is that the resulting representation is semantically richer and spatially more precise.
Interior Segmentation
A perfect Web application for decision boundary visualization.
Visual Display of Decision Boundary

Reviews

Changes saved
Showing 1 - 5 out of 24 reviews
Filter reviews by:
1.0
₹3,001.00 INR
He didnt complete the work even though he took the full amount for the project. Even before completing the work he forced to close the project on freelancers. After closing the project he is not responding to the messages.
M
Flag of GB Meghana H. @Meghanahc
3 months ago
4.8
₹11,000.00 INR
Very knowledgeable, gives his best to make the client understand the technical part and ready to help whenever needed. Just a suggestion that try to work a little on communication part that how to explain in a much better way for smooth conversation. Thank you so much for all the help!
Python
Machine Learning (ML)
Data Mining
OpenCV
+1 more
T
Flag of IN Tanishka D. @tanishka1125
4 months ago
5.0
$60.00 USD
great work and satisfied
Article Writing
Machine Learning (ML)
Natural Language
N
Flag of SA Nina F. @ninanina1990
5 months ago
5.0
$25.00 USD
Fast, great work and great help!
Python
Machine Learning (ML)
Data Mining
OpenCV
+1 more
G
Flag of ES Gopal L. @gopalrajora1
5 months ago
5.0
$70.00 USD
It was an awesome experience working with him. his work quality was very good and communication was clear. would like to work with him in future for sure.
Python
Machine Learning (ML)
Natural Language
Artificial Intelligence
Deep Learning
G
Flag of ES Gopal L. @gopalrajora1
5 months ago

Experience

Machine Learning Engineer

AWS
Sep 2019 - Present
I am working as a machine learning engineer here.

Testing automation engineer

Qualcomm
Apr 2018 - Aug 2019 (1 year, 4 months)
I worked on NLP for automizing the emailing system, studing the process logs and handling the client essue etc. I got a patent on Deep Dependency Parsing for QXDM log analyser.

Education

BTech

Indian Institute of Technology, Delhi, India 2014 - 2018
(4 years)

Qualifications

Face Manipulation Detection Challange.

Microsoft
2020
Here the problem statement was to design a model which can detect and segment manipulated face. Our rank was 13 among 1500 + teams.

IBM AI Hackathon

IBM
2019
My rank was 7.

Publications

Investigation on DeepLab segmentation edges

CVPR
The edge blurring issue of deeplab is resolved.

Contact Kaibalya B. about your job

Log in to discuss any details over chat.

Verifications

Preferred Freelancer
Identity Verified
Payment Verified
Phone Verified
Email Verified
Facebook Connected

Top Skills

Python
18
Machine Learning (ML)
16
Deep Learning
14
OpenCV
11
Data Mining
10

Browse Similar Freelancers

Python Developers in India
Python Developers
Machine Learning Experts
Deep Learning Specialists

Browse Similar Showcases

Python
Machine Learning (ML)
Deep Learning
OpenCV
Previous User
Next User
Invite sent successfully!
Registered Users Total Jobs Posted
Freelancer ® is a registered Trademark of Freelancer Technology Pty Limited (ACN 142 189 759)
Copyright © 2021 Freelancer Technology Pty Limited (ACN 142 189 759)
There is no internet connection
Loading preview