- 139
- 147 441
Polo Club of Data Science
United States
Registrace 3. 08. 2016
At Georgia Tech, our research group innovates at the intersection of data mining and human-computer interaction (HCI) to synthesize scalable, interactive, and interpretable tools that amplify human’s ability to understand and interact with billion-scale data and machine learning models. Our focus application areas include cybersecurity (e.g., fraud detection, malware detection), adversarial machine learning (how to protect and break AI and deep learning), health, energy and social good.
This CZcams Channel present video highlights of research from our group. For more details about our research project, visit our homepage at poloclub.gatech.edu
This CZcams Channel present video highlights of research from our group. For more details about our research project, visit our homepage at poloclub.gatech.edu
ARCollab: Mobile Augmented Reality for Cardiovascular SurgicalPlanning
ARCollab is a collaborative cardiovascular surgical planning iOS application in mobile augmented reality. Multiple users can join a shared session and view the a patient’s 3D heart model from different perspectives. ARCollab allows surgeons and cardiologists to collaboratively interact with a 3D heart model in real-time. Changes made on a device update the model’s orientation on the other devices in the session.
Paper: arxiv.org/abs/2402.05075
Code: github.com/poloclub/arcollab
Paper: arxiv.org/abs/2402.05075
Code: github.com/poloclub/arcollab
zhlédnutí: 36
Video
Demo Video "LLM Attributor: Interactive Visual Attribution for LLM Generation"
zhlédnutí 180Před měsícem
This is a demo video for the paper "LLM Attributor: Interactive Visual Attribution for LLM Generation" LLM Attributor helps you visualize training data attribution of text generation of your large language models (LLMs). Interactively select text phrases and visualize the training data points responsible for generating the selected phrases. Easily modify model-generated text and observe how you...
Transformer Explainer: Interactive Learning of Text-Generative Models
zhlédnutí 382Před měsícem
Transformer Explainer is an interactive visualization tool designed to help anyone learn how Transformer-based models like GPT work. It runs a live GPT-2 model right in your browser, allowing you to experiment with your own text and observe in real time how internal components and operations of the Transformer work together to predict the next tokens. Try Transformer Explainer live demo: polocl...
How to Select Objects with Same Color in Figma
zhlédnutí 133Před měsícem
How to select objects by color in Figma? In the example from the video tutorial, the letter "E" consists of light pink pixels and some dark pink shadow pixels. First, select all pixels (both light pink and dark pink). Then, click the "target" icon next to the light pink fill color to select only the light pink pixels. Similarly, click the "target" icon next to the dark pink fill color to select...
Robust Principles: Architectural Design Principles for Adversarially Robust CNNs
zhlédnutí 174Před 8 měsíci
00:16 Overview of 3 generalizable architectural principles for adversarially robust CNNs 00:48 1st principle: optimal width-depth ratios 01:19 2nd principle: convolutional over patchify stem stage 01:52 3rd principle: robust residual block design 02:48 Consistent robustness improvements & #1 on RobustBench CIFAR-10 leaderboard Accepted by BMVC'23 RobustBench leaderboard: robustbench.github.io/#...
Revamp: Automated Simulations of Adversarial Attacks on Arbitrary Objects in Realistic Scenes
zhlédnutí 141Před 9 měsíci
Deep Learning models, such as those used in an autonomous vehicle are vulnerable to adversarial attacks where an attacker could place an adversarial object in the environment, leading to mis-classification. Generating these adversarial objects in the digital space has been extensively studied, however successfully transferring these attacks from the digital realm to the physical realm has prove...
Random Forest Machine Learning Algorithm Explained
zhlédnutí 286Před 10 měsíci
This is a lecture video of the Data and Visual Analytics (CSE6242/CX4242) course at Georgia Tech. Course website and lecture slides: poloclub.github.io/#cse6242 CSE6242 wk13 1 1 bagging and random forests
VisGrader: Automatic Grading of D3 Visualizations
zhlédnutí 160Před 11 měsíci
Manually grading D3 data visualizations is a challenging endeavor, and is especially difficult for large classes with hundreds of students. Grading an interactive visualization requires a combination of interactive, quantitative, and qualitative evaluation that are conventionally done manually and are difficult to scale up as the visualization complexity, data size, and number of students incre...
Stable Diffusion Explained with Examples - Visualizing Text-to-Image Generation
zhlédnutí 2,2KPřed rokem
Diffusion Explainer is the first interactive visualization tool that explains how Stable Diffusion transforms text prompts into images. Diffusion Explainer tightly integrates a visual overview of Stable Diffusion’s complex components with detailed explanations of their underlying operations, enabling fluid transition between multiple levels of abstraction through animations and interactive elem...
Literature Review Made Easy with Visualization: Argo Scholar Tutorial with Captions
zhlédnutí 670Před rokem
Easily find and visualize relevant papers for your research with Argo Scholar, directly in your web browsers. Free and no installation needed. Argo Scholar website: poloclub.github.io/argo-scholar/
Visual Auditor: Interactive Visualization for Detection and Summarization of Model Biases
zhlédnutí 96Před rokem
IEEE VIS 2022 Short Paper Presentation
Demo Video "Dodrio: Exploring Transformer Models with Interactive Visualization"
zhlédnutí 774Před rokem
This is a demo video for the manuscript: "Dodrio: Exploring Transformer Models with Interactive Visualization" For a live demo, visit: poloclub.github.io/dodrio Music: Carefree by Cityfires Link: opengameart.org/content/acrostics License: Creative Commons 3.0
How to add text box to Tableau dashboard
zhlédnutí 16KPřed rokem
Drag a "Text" from the "Objects" side bar and release it at a location that you want on your dashboard.
HILDA 2022 Keynote: Visualization for Deep Learning: An Odyssey
zhlédnutí 317Před 2 lety
Machine learning and AI models are growing increasingly available, but many models offer predictions that are difficult to understand, evaluate and ultimately act upon. We share our research journey in developing interactive visualization tools that help amplify people’s ability to understand and interact with large-scale data and complex models, sampling from projects where interactive visuali...
Demo Video "VisCUIT: Visual Auditor for Bias in CNN Image Classifier"
zhlédnutí 517Před 2 lety
Demo Video "VisCUIT: Visual Auditor for Bias in CNN Image Classifier"
NeuroCartography: Scalable Automatic Visual Summarization of Concepts in Deep Neural Networks
zhlédnutí 640Před 2 lety
NeuroCartography: Scalable Automatic Visual Summarization of Concepts in Deep Neural Networks
Best of Both Worlds: Robust Accented Speech Recognition with Adversarial Transfer Learning
zhlédnutí 207Před 2 lety
Best of Both Worlds: Robust Accented Speech Recognition with Adversarial Transfer Learning
Towards Automatic Grading of D3.js Visualizations
zhlédnutí 241Před 2 lety
Towards Automatic Grading of D3.js Visualizations
Argo Scholar: Interactive Visual Exploration of Literature in Browsers
zhlédnutí 210Před 2 lety
Argo Scholar: Interactive Visual Exploration of Literature in Browsers
Demo Video "Dodrio: Exploring Transformer Models with Interactive Visualization"
zhlédnutí 4KPřed 3 lety
Demo Video "Dodrio: Exploring Transformer Models with Interactive Visualization"
Interactive Scalable Visualizations for Data Discoveries and Interpretable AI
zhlédnutí 330Před 3 lety
Interactive Scalable Visualizations for Data Discoveries and Interpretable AI
Study and Work in USA for International Students
zhlédnutí 307Před 3 lety
Study and Work in USA for International Students
CardiacAR: Interactive Cardiovascular Surgical Planning via Augmented Reality
zhlédnutí 387Před 3 lety
CardiacAR: Interactive Cardiovascular Surgical Planning via Augmented Reality
LinkedIn Data Science: Creating Global Economic Opportunities
zhlédnutí 1KPřed 3 lety
LinkedIn Data Science: Creating Global Economic Opportunities
SkeletonVis - Visualization for Understanding Adversarial Attacks on Human Action Recognition Models
zhlédnutí 346Před 3 lety
SkeletonVis - Visualization for Understanding Adversarial Attacks on Human Action Recognition Models
Bluff: Interactively Deciphering Adversarial Attacks on Deep Neural Networks
zhlédnutí 681Před 3 lety
Bluff: Interactively Deciphering Adversarial Attacks on Deep Neural Networks
A Comparative Analysis of Industry Human-AI Interaction Guidelines
zhlédnutí 630Před 3 lety
A Comparative Analysis of Industry Human-AI Interaction Guidelines
PeopleMap: Visualization Tool for Mapping Out Researchers using Natural Language Processing
zhlédnutí 172Před 3 lety
PeopleMap: Visualization Tool for Mapping Out Researchers using Natural Language Processing
Well done!
This lab always produces high quality research. Uniquely, they make very intuitive user friendly UIs for their concepts.
best explanation with awesome representation of the intuition in pictorial form
at 16:31, if you only interested in some nodes, then the (1-c)/n in the slide is not correct. It must be (1-c)/n', where n' is sizeof(some nodes).
thank you. Can we still be sure that a solution of p exists always ?
Brilliant! Thank you Dr Chau!
Thank you very much!
nice job
nice work
How can I find an internship? I'm one of your students in this course on DVA Fall 2022 Thank you Professor
Multiple possible ways: career fairs of your college and academic program, and importantly leverage you network of connections, such as alumni of your programs, peers who have interned before or are doing internships/co-ops currently, friends from your former schools, connections on LinkedIn, etc.
Thanks for sharing...it's a really helpful video.
WOW this is really amazing it’s like a grad-CAM but for nlp , it’s going to be super useful for understanding models when tuning models and experimenting with building different model architectures thank you for making this open source
Please, share links of your work for ease of access. Thanks
This is literally the greatest thing I've ever seen
perfect
Thank you!
Great job
Thank you Khaled!
Great work, i am currently also working on a novel attack for Action Recognition models,
Hello, good explanation. However, I still have a question, how do we know K? How is K computed?
Typically, when we set k to be smaller than the rank of matrix A, that means we're interested in only the k "most important" concepts. Exactly how many of such (k) "important" concepts to use will be determined by your use case.
I'm so confused
This is showing R-CNN, it's an algorithm that detects objects. They modified the second stop sign just enough so that the algorithm recognizes it as a person. The percentages you see are the "confidence" the program has that it identified the object correctly.
Can be a super useful tool for education on DNN and interpretable AI. Nice job!
This is super awesome.. great work.. 👍
Very cool!
how to get gamut ? gihub any links or etc source `?
msrgamut.microsoft.com/