2021 February

Fun

  • Pet Rock

Machine Learning: Blog

  • [Rahmadya Trias Handayanto "Just for a little kindness"](https://rahmadya.com/)

Machine Learning: Image Generator

Hardware: Project

Scientific Report Tools

Time Tracker and Productivity

Privacy

  • [Opinion They Stormed the Capitol. Their Apps Tracked Them. - The New York Times](https://www.nytimes.com/2021/02/05/opinion/capitol-attack-cellphone-data.html)

Search: Image

Visualization: Neural Network

Visualization: Complexity

  • Complexity Explained
  • [Complexity Explorables Complexity Explorables](https://www.complexity-explorables.org/)

Web-based Local Only Editor

Spartan Web Selfhosted

Android Machine Learning

Userscripts

Tunneling

Selfhosted

Video

Machine Learning Visualization

Sound: Ambient

Machine Learning: Action Recogntion

SVG Generator

  • [Vector Pattern Generator VISIWIG](https://www.visiwig.com/patterns/)
  • SvgPathEditor

SVG Images

Music Generation

Wikimedia

  • [Wikipedia Pageviews Analysis](https://pageviews.toolforge.org/?project=en.wikipedia.org&platform=all-access&agent=user&redirects=0&range=latest-20&pages=Cat Dog)

Python Tutorials

Python by Projects

Robot: Farming

  • [FarmBot Open-Source CNC Farming](https://farm.bot/)

Github

In-browser Object Detection

YOLO

There are a few different algorithms for object detection and they can be split into two groups:

  • Algorithms based on classification – they work in two stages. In the first step, we're selecting from the image interesting regions. Then we're classifying those regions using convolutional neural networks. This solution could be very slow because we have to run prediction for every selected region. Most known example of this type of algorithms is the Region-based convolutional neural network (RCNN) and their cousins Fast-RCNN and Faster-RCNN.
  • Algorithms based on regression – instead of selecting interesting parts of an image, we're predicting classes and bounding boxes for the whole image in one run of the algorithm. Most known example of this type of algorithms is YOLO (You only look once) commonly used for real-time object detection.

YOLO metrics:

In-Browser Pose Identification

Colorization

  • Pix2pix vs CycleGAN
    • Cycle-GAN can work in an ‘unpaired' manner and various architectural differences. Unpaired image translation is much harder, as you demand the model to learn objects of different shapes, size, angle, texture, location in different scenes and settings on top of the actual task (coloring in this case). Requires more data and you don't have fine control on the learning. Formulating the coloring problem as a paired task makes more sense as you simply decrease the complexity of the problem without increasing data collection/annotation work.
    • The whole point about using CycleGAN is that it can learn in unpaired situations. And it works well in the context of style transfer tasks where the changes are really bold and less nuanced. But, in the context of image colorization, the changes are really subtle and also there are way more options to choose colors than changing a horse to zebra. The other thing is that learning to change a colored image to black and white is much easier for the model than learning to colorize it which can lead to a bad learning procedure.
    • The most prominent differences is that CycleGAN helps when you have unpaired images and you want to go from one class to the other (Horse to Zebra for example) but in the Pix2Pix paper, the images that you get after the inference, are the input images but with some new features (black&white to colorized or day time to night time of a scene). In pix2pix, a conditional GAN (one generator and one discriminator) is used with some supervision from L1 loss. In CycleGAN, you need two generators and two discriminators to do the task: one pair for going from class A to B and one pair for going from class B to A. Also you need Cycle Consistency Loss to make sure that the models learn to undo the changes they make.
  • pix2pixhd is pix2pix in higher resolution
  • pix2pix tutorial and example:

Segmentation

  • One-Shot Medical Image Segmentation: All you need is one labeled image
  • [Dense cellular segmentation for EM using 2D–3D neural network ensembles Scientific Reports](https://www.nature.com/articles/s41598-021-81590-0)
    • [Dense cellular segmentation for EM using 2D-3D neural network ensembles leapmanlab.github.io](https://leapmanlab.github.io/dense-cell/)

Deep Learning : Review on Review

Timezone

Machine Learning Blog

Image Tools

  • [SVG Pattern Generator Create Seamless, Royalty-Free Patterns.](https://doodad.dev/pattern-generator/)
  • [Dither Me This Image Dithering Tool](https://doodad.dev/dither-me-this/)

eBook

NLP

Course

Map OSM

Animation

Rooftop Detection Machine Learning

Downloader

Video Editor

  • Shotcut - Home
  • [OpenShot Video Editor Free, Open, and Award-Winning Video Editor for Linux, Mac, and Windows!](https://www.openshot.org/)

Research

File Send

Notetaking

Single HTML Web

Machine Learning

Audio

Cell Tower Dataset

AI for Humanity

AI for Audio

Machine Learning Course

Visualization

Adversarial AI

  • [adversarial.io – Fighting mass image recognition](https://adversarial.io/)

Course

AI Blog

  • [Data Science Weekly Newsletter Archive Data Science Weekly](https://www.datascienceweekly.org/newsletters)

Gig Economy

  • Workers or self-employed?
  • The court made very specific arguments:
    1. Uber set the fare which meant that they dictated how much drivers could earn
    2. Uber set the contract terms and drivers had no say in them ‘
    3. Request for rides is constrained by Uber who can penalise drivers if they reject too many rides
    4. Uber monitors a driver's service through the star rating and has the capacity to terminate the relationship if after repeated warnings this does not improve
  • The main difference I see is the imbalance of power. A self-employed can decide who to work with and negotiate almost everything with them. He have my business, the company have theirs, and they discuss to try to reach agreements or not. Uber drivers are not in this position. Uber can dictate everything they want, drivers can only decide to stay or leave.
  • The issue here is that Uber pretends the driver is contracted by the passenger. In which case the driver should be free to negotiate with the passenger. That Uber exercises control of the relationship is what is indicating that Uber is an employer here. Uber pretends you're working for the customer, not them. That is a key element in their attempt to pretend they're not an employer.
  • Uber acts as an intermediary party between the contractor and the client (the passenger), so you should compare it to platforms for contract work, such as Upwork, or fiverr, or something like that. The point of Uber is that they're only an intermediate party that puts clients and drivers together. Drivers are not providing a service to Uber. They don't control your rates, they don't stop you from getting jobs if you reject a lot, they don't set any quality minimums for your job, and they don't remove you if someone doesn't give you five stars. In other words, you're not working for these platforms, it's just a marketplace. With Uber, drivers are working for Uber, not for the passengers. However, Uber is pretending to be a marketplace, and it isn't. This sentence just proves that.

Blog with AI

  • [fastpages An easy to use blogging platform with support for Jupyter Notebooks.](https://fastpages.fast.ai/)
  • [Data science notebooks in Review 2020 by Elizabeth Dlha Deepnote Deepnote](https://medium.com/deepnote/data-science-notebooks-2020-review-15a3c95cfc09)
  • [GuitarML Zak Jost](https://blog.zakjost.com/post/guitarml/) Audio Enhancement Machine Learning

Face Recognition

Social Media

Notebook

  • [The Collaborative Data Science Platform Mode](https://mode.com/)

ML Tools

  • [RapidMiner Best Data Science & Machine Learning Platform](https://rapidminer.com/)
    • [Machine Learning and RapidMiner Tutorials RapidMiner Academy](https://academy.rapidminer.com/)
    • [Online Certification RapidMiner Academy](https://academy.rapidminer.com/pages/certification)

Free Stuffs

Web Scraping dan Data Mining

AI Learning

  • [Data Science and Cognitive Computing Courses - Cognitive Class Free Courses in Data Science, AI, Cloud Computing, Containers, Kubernetes, Blockchain and more.](https://cognitiveclass.ai/)

NLP

Moodle

FPGA

Open Hardware

3D Reconstruction

Portable Server

Video Editor

FPGA

Twitter Dataset

Interesting Article

Fonts

Web Tools

Data Visualization