We suffered through curating and analysing thousands of benchmarks -- to better understand the (mis)measurement of AI! 📏🤖🔬

We cover all of #NLProc and #ComputerVision.

Now live at @NatureComms! nature.com/articles/s4146…

1/

We cover all of #NLProc and #ComputerVision.

Now live at @NatureComms! nature.com/articles/s4146…

1/

excited to release the course material (slides, lecture notes, coding+theory exercises, Unity-based simulator, ...) for the graduate course MIT 16.485 "Visual Navigation for Autonomous Vehicles" (VNAV), with topics spanning geometric vision, VIO, SLAM, 3D reconstruction, ...

2/n ... robust estimation, graph neural networks, and more.

Course material: vnav.mit.edu

Paper describing the course: arxiv.org/pdf/2206.00777…

Course material: vnav.mit.edu

Paper describing the course: arxiv.org/pdf/2206.00777…

3/n The course was created at MIT in 2018 in collaboration with @KasraKhosoussi, and with many contributors including @VassilisTzoumas , Golnaz Habibi,

@MarkusRyll , @TalakRajat , @jingnanshi , and Pasquale Antonante, among others.

@MarkusRyll , @TalakRajat , @jingnanshi , and Pasquale Antonante, among others.

This week @NVIDIA open sourced the 3D object generation AI model, GET3D. GET3D is a generative model of high quality 3D textured shapes learned from images.

1/4

1/4

Trained using only 2D images, GET3D generates 3D shapes with high-fidelity textures and complex geometric details.

2/4

2/4

These 3D objects are created in the same format used by popular graphics software applications, allowing users to immediately import their shapes into 3D renderers and game engines for further editing.

3/4

3/4

The First-Time Machine Learning Playbook.

(Read this if you want to efficiently learn machine learning, avoid frustration from searching for resources, and build a career in ML.)

#Ship30For30 #ML #MachineLearning

(Read this if you want to efficiently learn machine learning, avoid frustration from searching for resources, and build a career in ML.)

#Ship30For30 #ML #MachineLearning

What people think you need:

Most people think you need to be a math and stats expert to learn ML. ML can be math intensive, but it’s not a barrier to entry.

What you need:

• The Ability to code

• An Open and Curious Mind

• A Good starting point

Which leads me to …

Most people think you need to be a math and stats expert to learn ML. ML can be math intensive, but it’s not a barrier to entry.

What you need:

• The Ability to code

• An Open and Curious Mind

• A Good starting point

Which leads me to …

Where to Start:

If you can code, then start here: course.fast.ai

The @fastdotai Course is a fantastic place to start. Instead of going from the basics towards an application, it starts from an application and breaks it down piece by piece.

Did I mention it’s free?

If you can code, then start here: course.fast.ai

The @fastdotai Course is a fantastic place to start. Instead of going from the basics towards an application, it starts from an application and breaks it down piece by piece.

Did I mention it’s free?

Happy to finally share our paper about differentiable Top-K Learning by Sorting that didn’t make it to #CVPR2022, but was accepted for #ICML2022! We show that you can improve classification by actually considering top-1 + runner-ups… 1/6🧵

#ComputerVision #AI #MachineLearning

#ComputerVision #AI #MachineLearning

Paper: arxiv.org/abs/2206.07290

Great work by @FHKPetersen in collaboration with Christian Borgelt, @OliverDeussen . 2/6🧵

@MITIBMLab @goetheuni @UniKonstanz

Great work by @FHKPetersen in collaboration with Christian Borgelt, @OliverDeussen . 2/6🧵

@MITIBMLab @goetheuni @UniKonstanz

Idea: Top-k class accuracy is used in many ML tasks, but training is usually limited to top-1 accuracy (or another k). We propose a differentiable top-k classification loss that allows training by considering any combination of top-k predictions, e.g. top-2 top-5, 3/6🧵

Check out our #CVPR2022 paper! We improve multimodal zero-shot text-to-video retrieval on Youcook2/MSR-VTT by leveraging fusion transformer and combinatorial loss. 1/🧵

#ComputerVision #AI #MachineLearning

@MITIBMLab @goetheuni @MIT_CSAIL @IBMResearch

#ComputerVision #AI #MachineLearning

@MITIBMLab @goetheuni @MIT_CSAIL @IBMResearch

If you want to go directly to the paper/code, please check out:

paper: arxiv.org/abs/2112.04446

Github link: github.com/ninatu/everyth…

Great work by @ninashv__ , @Brian271828, @arouditchenko Samuel Thomas, Brian Kingsbury, @RogerioFeris , David Harwath, and James Glass.

paper: arxiv.org/abs/2112.04446

Github link: github.com/ninatu/everyth…

Great work by @ninashv__ , @Brian271828, @arouditchenko Samuel Thomas, Brian Kingsbury, @RogerioFeris , David Harwath, and James Glass.

1/ "Software is eating the world. Machine learning is eating software. Transformers are eating machine learning."

Let's understand what these Transformers are all about

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataAnalytics

Let's understand what these Transformers are all about

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataAnalytics

2/ #Transformers architecture follows Encoder and Decoder structure.

The encoder receives input sequence and creates intermediate representation by applying embedding and attention mechanism.

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI

The encoder receives input sequence and creates intermediate representation by applying embedding and attention mechanism.

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI

3/ Then, this intermediate representation or hidden state will pass through the decoder, and the decoder starts generating an output sequence.

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist #DataAnalytics #Statistics

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist #DataAnalytics #Statistics

But what p-value means in #MachineLearning - A thread

It tells you how likely it is that your data could have occurred under the null hypothesis

1/n

#DataScience #DeepLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat

It tells you how likely it is that your data could have occurred under the null hypothesis

1/n

#DataScience #DeepLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat

2/n

What Is a Null Hypothesis?

A null hypothesis is a type of statistical hypothesis that proposes that no statistical significance exists in a set of given observations.

#DataScience #MachineLearning #100DaysOfMLCode #Python #stat #Statistics #Data #AI #Math #deeplearning

What Is a Null Hypothesis?

A null hypothesis is a type of statistical hypothesis that proposes that no statistical significance exists in a set of given observations.

#DataScience #MachineLearning #100DaysOfMLCode #Python #stat #Statistics #Data #AI #Math #deeplearning

3/n

A P-value is the probability of obtaining an effect at least as extreme as the one in your sample data, assuming the truth of the null hypothesis

#DataScience #MachineLearning #100DaysOfMLCode #Python #DataScientist #Statistics #Data #DataAnalytics #AI #Math

A P-value is the probability of obtaining an effect at least as extreme as the one in your sample data, assuming the truth of the null hypothesis

#DataScience #MachineLearning #100DaysOfMLCode #Python #DataScientist #Statistics #Data #DataAnalytics #AI #Math

1/ One way to test whether a time series is stationary is to perform an augmented Dickey-Fuller test - A Thread

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist #DataAnalytics #Statistics #programming #ArtificialIntelligence

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist #DataAnalytics #Statistics #programming #ArtificialIntelligence

2/ H0: The time series is non-stationary. In other words, it has some time-dependent structure and does not have constant variance over time.

HA: The time series is stationary.

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist

HA: The time series is stationary.

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist

3/ If the p-value from the test is less than some significance level (e.g. α = .05), then we can reject the null hypothesis and conclude that the time series is stationary.

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist

Kullback-Leibler (KL) Divergence - A Thread

It is a measure of how one probability distribution diverges from another expected probability distribution.

#DataScience #Statistics #DeepLearning #ComputerVision #100DaysOfMLCode #Python #programming #ArtificialIntelligence #Data

It is a measure of how one probability distribution diverges from another expected probability distribution.

#DataScience #Statistics #DeepLearning #ComputerVision #100DaysOfMLCode #Python #programming #ArtificialIntelligence #Data

1/ When it is important to standardize variables in #DataScience #MachineLearning ? - A Thread

#DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist #DataAnalytics #Statistics #programming #ArtificialIntelligence #Data #Stats #Database #BigData #100DaysOfCode

#DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist #DataAnalytics #Statistics #programming #ArtificialIntelligence #Data #Stats #Database #BigData #100DaysOfCode

2/ It is important to standardize variables before running Cluster Analysis. It is because cluster analysis techniques depend on the concept of measuring the distance between the different observations we're trying to cluster.

#DataScience #MachineLearning #DeepLearning

#DataScience #MachineLearning #DeepLearning

3/ If a variable is measured at a higher scale than the other variables, then whatever measure we use will be overly influenced by that variable.

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist #DataAnalytics #Statistics

#DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #pythoncode #AI #DataScientist #DataAnalytics #Statistics

Did you know how TensorFlow can run on a single mobile device as well as on an entire data center? Read this thread

1/n

#TensorFlow #DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data

1/n

#TensorFlow #DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data

2/n

Google has designed TensorFlow such that it is capable of dividing a large model graph whenever needed.

#TensorFlow #DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat #AI

Google has designed TensorFlow such that it is capable of dividing a large model graph whenever needed.

#TensorFlow #DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat #AI

3/n

It assigns special SEND and RECV nodes whenever a graph is divided between multiple devices (CPUs or GPUs).

#TensorFlow #DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat #AI

It assigns special SEND and RECV nodes whenever a graph is divided between multiple devices (CPUs or GPUs).

#TensorFlow #DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat #AI

Outlier Detection with Alibi Detect Library - A Thread

1/n

#DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat #pythoncode #AI #ArtificialIntelligence

1/n

#DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat #pythoncode #AI #ArtificialIntelligence

2/n

Alibi Detect is a Python library for detecting outliers, adversarial data, and drift. Accommodates tabular data, text, images, and time series that can be used both online and offline. Both TensorFlow and PyTorch backends are supported

#DataScience #DeepLearning

Alibi Detect is a Python library for detecting outliers, adversarial data, and drift. Accommodates tabular data, text, images, and time series that can be used both online and offline. Both TensorFlow and PyTorch backends are supported

#DataScience #DeepLearning

3/n

Supports a variety of outlier detection techniques, including Mahalanobis distance, Isolation forest, and Seq2seq

#DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat #pythoncode

Supports a variety of outlier detection techniques, including Mahalanobis distance, Isolation forest, and Seq2seq

#DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat #pythoncode

1/ Can you classify something without seeing it before - that's what Zero-Shot Learning is all about - A Thread

👉 One of the popular methods for zero-shot learning is Natural Language Inference (NLI).

#DataScience #DeepLearning #MachineLearning #100DaysOfMLCode #Pytho

👉 One of the popular methods for zero-shot learning is Natural Language Inference (NLI).

#DataScience #DeepLearning #MachineLearning #100DaysOfMLCode #Pytho

2/ Zero-shot learning is when we test a model on a task for which it hasn't been trained.

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence #Data #pythoncode #AI #Stats #DeepLearning #100DaysOfCode #Database #BigData

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence #Data #pythoncode #AI #Stats #DeepLearning #100DaysOfCode #Database #BigData

3/ In Zero-shot classification, we ask the model to classify a sentence to one of the classes (label) that the model hasn't seen during training.

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence #Data #pythoncode #AI

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence #Data #pythoncode #AI

What is p-value - A thread

It tells you how likely it is that your data could have occurred under the null hypothesis.

1/n

#DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat

It tells you how likely it is that your data could have occurred under the null hypothesis.

1/n

#DataScience #DeepLearning #MachineLearning #ComputerVision #100DaysOfMLCode #Python #DataScientist #Statistics #programming #Data #Math #Stat

2/n

What Is a Null Hypothesis?

A null hypothesis is a type of statistical hypothesis that proposes that no statistical significance exists in a set of given observations.

#DataScience #MachineLearning #100DaysOfMLCode #Python #stat #Statistics #Data #AI #Math #deeplearning

What Is a Null Hypothesis?

A null hypothesis is a type of statistical hypothesis that proposes that no statistical significance exists in a set of given observations.

#DataScience #MachineLearning #100DaysOfMLCode #Python #stat #Statistics #Data #AI #Math #deeplearning

3/n

A P-value is the probability of obtaining an effect at least as extreme as the one in your sample data, assuming the truth of the null hypothesis

#DataScience #MachineLearning #100DaysOfMLCode #Python #DataScientist #Statistics #Data #DataAnalytics #AI #Math

A P-value is the probability of obtaining an effect at least as extreme as the one in your sample data, assuming the truth of the null hypothesis

#DataScience #MachineLearning #100DaysOfMLCode #Python #DataScientist #Statistics #Data #DataAnalytics #AI #Math

Some tips to increase the Performance of Your Object Detection Models - A trhead

1/n

#computervision #pytorch #deeplearning #deeplearningai #100daysofmlcode #neuralnetworks #machinelearning #datascience #pythonprogramming #python #AI #ArtificialIntelligence #pythonprogramming

1/n

#computervision #pytorch #deeplearning #deeplearningai #100daysofmlcode #neuralnetworks #machinelearning #datascience #pythonprogramming #python #AI #ArtificialIntelligence #pythonprogramming

2/n

Following tips may boost model performance across different network structures with up to 5% (mAP or mean Average Precision) without increasing computational costs in any way.

#computervision #pytorch #deeplearning #deeplearningai #100daysofmlcode #neuralnetworks #AI

Following tips may boost model performance across different network structures with up to 5% (mAP or mean Average Precision) without increasing computational costs in any way.

#computervision #pytorch #deeplearning #deeplearningai #100daysofmlcode #neuralnetworks #AI

3/n

Visually Coherent Image Mix-up for Object Detection. This has already been proven to be successful in lessening adversarial fears in network classification after testing it on COCO 2017 and PASCAL datasets with YOLOv3 models.

#computervision #pytorch

Visually Coherent Image Mix-up for Object Detection. This has already been proven to be successful in lessening adversarial fears in network classification after testing it on COCO 2017 and PASCAL datasets with YOLOv3 models.

#computervision #pytorch

🤖 Optical Character Recognition (OCR) is one of the most important applications of #ComputerVision in the real world.

In this thread, we cover some of our popular free tutorials on #OCR, which will get you started in no time. 🚀

🧵👇

In this thread, we cover some of our popular free tutorials on #OCR, which will get you started in no time. 🚀

🧵👇

What is Optical Character Recognition?

✨Accept an image

✨Detect the text

✨Convert the text to machine-readable format

🔗pyimagesearch.com/2021/08/09/wha…

[1/N]

✨Accept an image

✨Detect the text

✨Convert the text to machine-readable format

🔗pyimagesearch.com/2021/08/09/wha…

[1/N]

Installing #Tesseract, #PyTesseract, and #Python OCR Packages On Your System

✨Install the Tesseract OCR engine on your machine

✨Create a python virtual environment for installation

✨Install necessary python packages

🔗pyimagesearch.com/2021/08/16/ins…

[2/N]

✨Install the Tesseract OCR engine on your machine

✨Create a python virtual environment for installation

✨Install necessary python packages

🔗pyimagesearch.com/2021/08/16/ins…

[2/N]

#Highlights2021 for me: our #survey on efficient processing of #sparse and compressed tensors of #ML/#DNN models on #hardware accelerators published in @ProceedingsIEEE.

Paper: dx.doi.org/10.1109/JPROC.…

arXiv: arxiv.org/abs/2007.00864

RT/sharing appreciated. 🧵

Paper: dx.doi.org/10.1109/JPROC.…

arXiv: arxiv.org/abs/2007.00864

RT/sharing appreciated. 🧵

Context: Tensors of ML/DNN are compressed by leveraging #sparsity, #quantization, shape reduction. We summarize several such sources of sparsity & compression (§3). Sparsity is induced in structure while pruning & it is unstructured inherently for various applications or sources.

Likewise, leveraging value similarity or approximate operations could yield irregularity in processing. Also, techniques for size-reduction make tensors asymmetric-shaped. Hence, special mechanisms can be required for efficient processing of sparse and irregular computations.

Incredibly happy that our @RoboStack paper has been accepted to the @ieeeras Robotics & Automation Magazine 🥳. @RoboStack brings together #ROS @rosorg with @condaforge and @ProjectJupyter. Preprint: arxiv.org/abs/2104.12910. Find out some key benefits in this 🧵: 1/n

You can now easily use ROS (both ROS1 #Noetic and ROS2 #Galactic) on a wide range of platforms: @linuxfoundation (not just a specific Ubuntu, any Linux!), @Apple #MacOS and @Microsoft @Windows - even on ARM processors including the new M1 (work still in progress, though). 2/n

Thanks to the tight coupling to @condaforge, this enables you to (very easily) install ROS side-by-side with thousands of scientific libraries, including recent #computervision and #machinelearning ones (think @TensorFlow, @opencvlibrary, @PyTorch and more). 3/n

@amaarora really explained #convolutions very well in #fastbook week 12 session which can be viewed here

I wasn't able to write a blog post explaining my learnings from the stream but would threfore write a 🧵

I wasn't able to write a blog post explaining my learnings from the stream but would threfore write a 🧵

2/n

After going through 1st part of #convolutions chapter, have cleared a concept and was introduced to two new concepts.

1. How depthwise convolutions work (3/n)

2. Dilated convolutions (7/n)

3. Alternate interpretation of #stride (9/n)

After going through 1st part of #convolutions chapter, have cleared a concept and was introduced to two new concepts.

1. How depthwise convolutions work (3/n)

2. Dilated convolutions (7/n)

3. Alternate interpretation of #stride (9/n)

3/n

When we have a n-channel input and a m-channel output, we need to convolve over not only 2-Dimensions (W x H) but also across the depth D.

An RGB image for example has 3 channels

Let us consider we want to derive 10 feature maps from this input.

When we have a n-channel input and a m-channel output, we need to convolve over not only 2-Dimensions (W x H) but also across the depth D.

An RGB image for example has 3 channels

Let us consider we want to derive 10 feature maps from this input.

"torch.manual_seed(3407) is all you need"!

draft 📜: davidpicard.github.io/pdf/lucky_seed…

Sorry for the title. I promise it's not (entirely) just for trolling. It's my little spare time project of this summer to investigate unaccounted randomness in #ComputerVision and #DeepLearning.

🧵👇 1/n

draft 📜: davidpicard.github.io/pdf/lucky_seed…

Sorry for the title. I promise it's not (entirely) just for trolling. It's my little spare time project of this summer to investigate unaccounted randomness in #ComputerVision and #DeepLearning.

🧵👇 1/n

The idea is simple: after years of reviewing deep learning stuff, I am frustrated of never seeing a paragraph that shows how robust the results are w.r.t the randomness (initial weights, batch composition, etc). 2/n

After seeing several videos by @skdh about how experimental physics claims tend to disappear through repetition, I got the idea of gauging the influence of randomness by scanning a large amount of seeds. 3/n

Quick Tweet Storm ⛈

How does AI bounding box detection work?

🧠 Learn in 30 seconds

#100DaysOfCode #CodeNewbie #MadeWithTFJS #MachineLearning #ComputerVision

How does AI bounding box detection work?

🧠 Learn in 30 seconds

#100DaysOfCode #CodeNewbie #MadeWithTFJS #MachineLearning #ComputerVision

It looks so simple when #AI does it right?

But #machinelearning doesn't give you an image, it gives you data. It's up to you to make it look simple.

But #machinelearning doesn't give you an image, it gives you data. It's up to you to make it look simple.

I'm really looking forward to participating in the forthcoming @markcubanai Boot Camps this Fall! I will be participating as a mentor and collaborating with @Caltech and the @AppAcademyPHS! cc @mcuban #AI #artificialintelligence #machinelearning #ML

The Mark Cuban Foundation works with local companies to host Introduction to #AI bootcamps for underserved high school (9th-12th) grade students at no cost.

the program does not have any pre-requisites or require any prior experience with coding. Students with any level of interest in technology will walk away from the @markcubanai program with a greater understanding of #AI

#CVPR2021 #cvpr2021_cv4aec is Live!

@li_fuxin is opening the 1st "Workshop & Challenge on #ComputerVision in the #BuiltEnvironment for the #Design, #Construction and #Operation of #Buildings"

@li_fuxin is opening the 1st "Workshop & Challenge on #ComputerVision in the #BuiltEnvironment for the #Design, #Construction and #Operation of #Buildings"

The winner of the 2D challenge is the Institute of Automation, Chinese Academy of Sciences

The winner of the 3D challenge is Purdue University

Congrats! #CVPR2021 #cvpr2021_cv4aec

The winner of the 3D challenge is Purdue University

Congrats! #CVPR2021 #cvpr2021_cv4aec