Michio JP

Michio JP

1630771176

Visual Odometry Techniques for Embodied PointGoal Navigation

PointNav-VO

The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation

Project Page | Paper

Table of Contents

  • Setup
  • Reproduction
  • Plug-and-play
  • Train
  • Citation

Setup

Install Dependencies

conda env create -f environment.yml

Install Habitat

The repo is tested under the following commits of habitat-lab and habitat-sim.

habitat-lab == d0db1b55be57abbacc5563dca2ca14654c545552
habitat-sim == 020041d75eaf3c70378a9ed0774b5c67b9d3ce99

Note, to align with Habitat Challenge 2020 settings (see Step 36 in the Dockerfile), when installing habitat-sim, we compiled without CUDA support as

python setup.py install --headless

There was a discrepancy between noises models in CPU and CPU versions which has now been fixed, see this issue. Therefore, to reproduce the results in the paper with our pre-trained weights, you need to use noises model of CPU-version.

Download Data

We need two datasets to enable running of this repo:

  1. Gibson scene dataset
  2. PointGoal Navigation splits, we need pointnav_gibson_v2.zip.

Please follow Habitat's instruction to download them. We assume all data is put under ./dataset with structure:

.
+-- dataset
|  +-- Gibson
|  |  +-- gibson
|  |  |  +-- Adrian.glb
|  |  |  +-- Adrian.navmesh
|  |  |  ...
|  +-- habitat_datasets
|  |  +-- pointnav
|  |  |  +-- gibson
|  |  |  |  +-- v2
|  |  |  |  |  +-- train
|  |  |  |  |  +-- val
|  |  |  |  |  +-- valmini

Reproduce

Download pretrained checkpoints of RL navigation policy and VO from this link. Put them under pretrained_ckpts with the following structure:

.
+-- pretrained_ckpts
|  +-- rl
|  |  +-- no_tune
|  |  |  +-- rl_no_tune.pth
|  |  +-- tune_vo
|  |  |  +-- rl_tune_vo.pth
|  +-- vo
|  |  +-- act_forward.pth
|  |  +-- act_left_right_inv_joint.pth

Run the following command to reproduce navigation results. On Intel(R) Xeon(R) CPU E5-2683 v4 @ 2.10GHz and a Nvidia GeForce GTX 1080 Ti, it takes around 4.5 hours to complete evaluation on all 994 episodes with navigation policy tuned with VO.

cd /path/to/this/repo
export POINTNAV_VO_ROOT=$PWD

export NUMBA_NUM_THREADS=1 && \
export NUMBA_THREADING_LAYER=workqueue && \
conda activate pointnav-vo && \
python ${POINTNAV_VO_ROOT}/launch.py \
--repo-path ${POINTNAV_VO_ROOT} \
--n_gpus 1 \
--task-type rl \
--noise 1 \
--run-type eval \
--addr 127.0.1.1 \
--port 8338

Use VO as a Drop-in Module

We provide a class BaseRLTrainerWithVO that contains all necessary functions to compute odometry in base_trainer_with_vo.py. Specifically, you can use _compute_local_delta_states_from_vo to compute odometry based on adjacent observations. The code sturcture will be something like:

local_delta_states = _compute_local_delta_states_from_vo(prev_obs, cur_obs, action)
cur_goal = compute_goal_pos(prev_goal, local_delta_states)

To get more sense about how to use this function, please refer to challenge2020_agent.py, which is the agent we used in HabitatChallenge 2020.

Train Your Own VO

See details in TRAIN.md

Citation

Please cite the following papers if you found our model useful. Thanks!

Xiaoming Zhao, Harsh Agrawal, Dhruv Batra, and Alexander Schwing. The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation. ICCV 2021.

@inproceedings{ZhaoICCV2021,
  title={{The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation}},
  author={Xiaoming Zhao and Harsh Agrawal and Dhruv Batra and Alexander Schwing},
  booktitle={Proc. ICCV},
  year={2021},
}

Download Details:
 

Author: Xiaoming-Zhao
Download Link: Download The Source Code
Official Website: https://github.com/Xiaoming-Zhao/PointNav-VO 
License: Apache-2.0

What is GEEK

Buddha Community

Corey Brooks

Corey Brooks

1657254050

Top 9+ Common CSS Mistakes To Avoid

In this tutorial, we'll summarise what the top 9+ CSS mistakes are and how to avoid them.

Top 9+ Common CSS Mistakes To Avoid

It’s easy to get tripped up with CSS. Here are some common CSS mistakes we all make.

1. Not Using a Proper CSS Reset

Web browsers are our fickle friends. Their inconsistencies can make any developer want to tear their hair out. But at the end of the day, they’re what will present your website, so you better do what you have to do to please them.

One of the sillier things browsers do is provide default styling for HTML elements. I suppose you can’t really blame them: what if a “webmaster” chose not to style their page? There has to be a fallback mechanism for people who choose not to use CSS.

In any case, there’s rarely a case of two browsers providing identical default styling, so the only real way to make sure your styles are effective is to use a CSS reset. What a CSS reset entails is resetting (or, rather, setting) all the styles of all the HTML elements to a predictable baseline value. The beauty of this is that once you include a CSS reset effectively, you can style all the elements on your page as if they were all the same to start with.

It’s a blank slate, really. There are many CSS reset codebases on the web that you can incorporate into your work. I personally use a modified version of the popular Eric Meyer reset and Six Revisions uses a modified version of YUI Reset CSS.

You can also build your own reset if you think it would work better. What many of us do is utilizing a simple universal selector margin/padding reset.

* { margin:0; padding:0; } 

Though this works, it’s not a full reset.

You also need to reset, for example, borders, underlines, and colors of elements like list items, links, and tables so that you don’t run into unexpected inconsistencies between web browsers. Learn more about resetting your styles via this guide: Resetting Your Styles with CSS Reset.

2. Over-Qualifying Selectors

Being overly specific when selecting elements to style is not good practice. The following selector is a perfect example of what I’m talking about:

ul#navigation li a { ... } 

Typically the structure of a primary navigation list is a <ul> (usually with an ID like #nav or #navigation) then a few list items (<li>) inside of it, each with its own <a> tag inside it that links to other pages.

This HTML structure is perfectly correct, but the CSS selector is really what I’m worried about. First things first: There’s no reason for the ul before #navigation as an ID is already the most specific selector. Also, you don’t have to put li in the selector syntax because all the a elements inside the navigation are inside list items, so there’s no reason for that bit of specificity.

Thus, you can condense that selector as:

#navigation a { ... } 

This is an overly simplistic example because you might have nested list items that you want to style differently (i.e. #navigation li a is different from #navigation li ul li a); but if you don’t, then there’s no need for the excessive specificity.

I also want to talk about the need for an ID in this situation. Let’s assume for a minute that this navigation list is inside a header div (#header). Let us also assume that you will have no other unordered list in the header besides the navigation list.

If that is the case, we can even remove the ID from the unordered list in our HTML markup, and then we can select it in CSS as such:

#header ul a { ... } 

Here’s what I want you to take away from this example: Always write your CSS selectors with the very minimum level of specificity necessary for it to work. Including all that extra fluff may make it look more safe and precise, but when it comes to CSS selectors, there are only two levels of specificity: specific, and not specific enough.

3. Not Using Shorthand Properties

Take a look at the following property list:

#selector { margin-top: 50px; margin-right: 0; margin-bottom: 50px; margin-left 0; }

What is wrong with this picture? I hope that alarm bells are ringing in your head as you notice how much we’re repeating ourselves. Fortunately, there is a solution, and it’s using CSS shorthand properties.

The following has the same effect as the above style declaration, but we’ve reduced our code by three lines.

#selector { margin: 50px 0; }

Check out this list of properties that deals with font styles:

font-family: Helvetica; font-size: 14px; font-weight: bold; line-height: 1.5;

We can condense all that into one line:

font: bold 14px/1.5 Helvetica; 

We can also do this for background properties. The following:

background-image: url(background.png); background-repeat: repeat-y; background-position: center top;

Can be written in shorthand CSS as such:

background: url(background.png) repeat-y center top; 

4. Using 0px instead of 0

Say you want to add a 20px margin to the bottom of an element. You might use something like this:

#selector { margin: 20px 0px 20px 0px; } 

Don’t. This is excessive.

There’s no need to include the px after 0. While this may seem like I’m nitpicking and that it may not seem like much, when you’re working with a huge file, removing all those superfluous px can reduce the size of your file (which is never a bad thing).

5. Using Color Names Instead of Hexadecimal

Declaring red for color values is the lazy man’s #FF0000. By saying:

color: red;

You’re essentially saying that the browser should display what it thinks red is. If you’ve learned anything from making stuff function correctly in all browsers — and the hours of frustration you’ve accumulated because of a stupid list-bullet misalignment that can only be seen in IE7 — it’s that you should never let the browser decide how to display your web pages.

Instead, you should go to the effort to find the actual hex value for the color you’re trying to use. That way, you can make sure it’s the same color displayed across all browsers. You can use a color cheatsheet that provides a preview and the hex value of a color.

This may seem trivial, but when it comes to CSS, it’s the tiny things that often lead to the big gotchas.

6. Redundant Selectors

My process for writing styles is to start with all the typography, and then work on the structure, and finally on styling all the colors and backgrounds. That’s what works for me. Since I don’t focus on just one element at a time, I commonly find myself accidentally typing out a redundant style declaration.

I always do a final check after I’m done so that I can make sure that I haven’t repeated any selectors; and if I have, I’ll merge them. This sort of mistake is fine to make while you’re developing, but just try to make sure they don’t make it into production.

7. Redundant Properties

Similar to the one above, I often find myself having to apply the same properties to multiple selectors. This could be styling an <h5> in the header to look exactly like the <h6> in the footer, making the <pre>‘s and <blockquote>‘s the same size, or any number of things in between. In the final review of my CSS, I will look to make sure that I haven’t repeated too many properties.

For example, if I see two selectors doing the same thing, such as this:

#selector-1 { font-style: italic; color: #e7e7e7; margin: 5px; padding: 20px } .selector-2 { font-style: italic; color: #e7e7e7; margin: 5px; padding: 20px }

I will combine them, with the selectors separated by a comma (,):

#selector-1, .selector-2 { font-style: italic; color: #e7e7e7; margin: 5px; padding: 20px }

I hope you’re seeing the trend here: Try to be as terse and as efficient as possible. It pays dividends in maintenance time and page-load speed.

8. Not Providing Fallback Fonts

In a perfect world, every computer would always have every font you would ever want to use installed. Unfortunately, we don’t live in a perfect world. @font-face aside, web designers are pretty much limited to the few so called web-safe fonts (e.g.

Arial, Georgia, serif, etc.). There is a plus side, though. You can still use fonts like Helvetica that aren’t necessarily installed on every computer.

The secret lies in font stacks. Font stacks are a way for developers to provide fallback fonts for the browser to display if the user doesn’t have the preferred font installed. For example:

#selector { font-family: Helvetica; }

Can be expanded with fallback fonts as such:

#selector { font-family: Helvetica, Arial, sans-serif; }

Now, if the user doesn’t have Helvetica, they can see your site in Arial, and if that doesn’t work, it’ll just default to any sans-serif font installed.

By defining fallback fonts, you gain more control as to how your web pages are rendered.

9. Unnecessary Whitespace

When it comes to trying to reduce your CSS file sizes for performance, every space counts. When you’re developing, it’s OK to format your code in the way that you’re comfortable with. However, there is absolutely no reason not to take out excess characters (a process known as minification) when you actually push your project onto the web where the size of your files really counts.

Too many developers simply don’t minify their files before launching their websites, and I think that’s a huge mistake. Although it may not feel like it makes much of a difference, when you have huge CSS files

10. Not Organizing Your CSS in a Logical Way

When you’re writing CSS, do yourself a favor and organize your code. Through comments, you can insure that the next time you come to make a change to a file you’ll still be able to navigate it. 

I personally like to organize my styles by how the HTML that I’m styling is structured. This means that I have comments that distinguish the header, body, sidebar, and footer. A common CSS-authoring mistake I see is people just writing up their styles as soon as they think of them.

The next time you try to change something and can’t find the style declaration, you’ll be silently cursing yourself for not organizing your CSS well enough.

11. Using Only One Stylesheet for Everything

This one’s subjective, so bear with me while I give you my perspective. I am of the belief, as are others, that it is better to split stylesheets into a few different ones for big sites for easier maintenance and for better modularity. Maybe I’ll have one for a CSS reset, one for IE-specific fixes, and so on.

By organizing CSS into disparate stylesheets, I’ll know immediately where to find a style I want to change. You can do this by importing all the stylesheets into a stylesheet like so:

@import url("reset.css"); @import url("ie.css"); @import url("typography.css"); @import url("layout.css"); 

Let me stress, however, that this is what works for me and many other developers. You may prefer to squeeze them all in one file, and that’s okay; there’s nothing wrong with that.

But if you’re having a hard time maintaining a single file, try splitting your CSS up.

12. Not Providing a Print Stylesheet

In order to style your site on pages that will be printed, all you have to do is utilize and include a print stylesheet. It’s as easy as:

<link rel="stylesheet" href="print.css" media="print" /> 

Using a stylesheet for print allows you to hide elements you don’t want printed (such as your navigation menu), reset the background color to white, provide alternative typography for paragraphs so that it’s better suited on a piece of paper, and so forth. The important thing is that you think about how your page will look when printed.

Too many people just don’t think about it, so their sites will simply print the same way you see them on the screen.


I Made These 2 BEGINNER CSS Mistakes

No matter how long you've been writing code, it's always a good time to revisit the basics. While working on a project the other day, I made 2 beginner mistakes with the CSS I was writing. I misunderstood both CSS specificity and how transform:scale affects the DOM!

Stack Overflow about transform:scale - https://stackoverflow.com/questions/32835144/css-transform-scale-does-not-change-dom-size 
CSS Specificity - https://www.w3schools.com/css/css_specificity.asp 

#css 

Arvel  Parker

Arvel Parker

1591177440

Visual Analytics and Advanced Data Visualization

Visual Analytics is the scientific visualization to emerge an idea to present data in such a way so that it could be easily determined by anyone.

It gives an idea to the human mind to directly interact with interactive visuals which could help in making decisions easy and fast.

Visual Analytics basically breaks the complex data in a simple way.

The human brain is fast and is built to process things faster. So Data visualization provides its way to make things easy for students, researchers, mathematicians, scientists e

#blogs #data visualization #business analytics #data visualization techniques #visual analytics #visualizing ml models

Michio JP

Michio JP

1630771176

Visual Odometry Techniques for Embodied PointGoal Navigation

PointNav-VO

The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation

Project Page | Paper

Table of Contents

  • Setup
  • Reproduction
  • Plug-and-play
  • Train
  • Citation

Setup

Install Dependencies

conda env create -f environment.yml

Install Habitat

The repo is tested under the following commits of habitat-lab and habitat-sim.

habitat-lab == d0db1b55be57abbacc5563dca2ca14654c545552
habitat-sim == 020041d75eaf3c70378a9ed0774b5c67b9d3ce99

Note, to align with Habitat Challenge 2020 settings (see Step 36 in the Dockerfile), when installing habitat-sim, we compiled without CUDA support as

python setup.py install --headless

There was a discrepancy between noises models in CPU and CPU versions which has now been fixed, see this issue. Therefore, to reproduce the results in the paper with our pre-trained weights, you need to use noises model of CPU-version.

Download Data

We need two datasets to enable running of this repo:

  1. Gibson scene dataset
  2. PointGoal Navigation splits, we need pointnav_gibson_v2.zip.

Please follow Habitat's instruction to download them. We assume all data is put under ./dataset with structure:

.
+-- dataset
|  +-- Gibson
|  |  +-- gibson
|  |  |  +-- Adrian.glb
|  |  |  +-- Adrian.navmesh
|  |  |  ...
|  +-- habitat_datasets
|  |  +-- pointnav
|  |  |  +-- gibson
|  |  |  |  +-- v2
|  |  |  |  |  +-- train
|  |  |  |  |  +-- val
|  |  |  |  |  +-- valmini

Reproduce

Download pretrained checkpoints of RL navigation policy and VO from this link. Put them under pretrained_ckpts with the following structure:

.
+-- pretrained_ckpts
|  +-- rl
|  |  +-- no_tune
|  |  |  +-- rl_no_tune.pth
|  |  +-- tune_vo
|  |  |  +-- rl_tune_vo.pth
|  +-- vo
|  |  +-- act_forward.pth
|  |  +-- act_left_right_inv_joint.pth

Run the following command to reproduce navigation results. On Intel(R) Xeon(R) CPU E5-2683 v4 @ 2.10GHz and a Nvidia GeForce GTX 1080 Ti, it takes around 4.5 hours to complete evaluation on all 994 episodes with navigation policy tuned with VO.

cd /path/to/this/repo
export POINTNAV_VO_ROOT=$PWD

export NUMBA_NUM_THREADS=1 && \
export NUMBA_THREADING_LAYER=workqueue && \
conda activate pointnav-vo && \
python ${POINTNAV_VO_ROOT}/launch.py \
--repo-path ${POINTNAV_VO_ROOT} \
--n_gpus 1 \
--task-type rl \
--noise 1 \
--run-type eval \
--addr 127.0.1.1 \
--port 8338

Use VO as a Drop-in Module

We provide a class BaseRLTrainerWithVO that contains all necessary functions to compute odometry in base_trainer_with_vo.py. Specifically, you can use _compute_local_delta_states_from_vo to compute odometry based on adjacent observations. The code sturcture will be something like:

local_delta_states = _compute_local_delta_states_from_vo(prev_obs, cur_obs, action)
cur_goal = compute_goal_pos(prev_goal, local_delta_states)

To get more sense about how to use this function, please refer to challenge2020_agent.py, which is the agent we used in HabitatChallenge 2020.

Train Your Own VO

See details in TRAIN.md

Citation

Please cite the following papers if you found our model useful. Thanks!

Xiaoming Zhao, Harsh Agrawal, Dhruv Batra, and Alexander Schwing. The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation. ICCV 2021.

@inproceedings{ZhaoICCV2021,
  title={{The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation}},
  author={Xiaoming Zhao and Harsh Agrawal and Dhruv Batra and Alexander Schwing},
  booktitle={Proc. ICCV},
  year={2021},
}

Download Details:
 

Author: Xiaoming-Zhao
Download Link: Download The Source Code
Official Website: https://github.com/Xiaoming-Zhao/PointNav-VO 
License: Apache-2.0

Arvel  Parker

Arvel Parker

1591184760

Visual Analytics Services for Data-Driven Decision Making

Visual analytics is the process of collecting, examining complex and large data sets (structured or unstructured) to get useful information to draw conclusions about the datasets and visualize the data or information in the form of interactive visual interfaces and graphical manner.

Data analytics is usually accomplished by extracting or collecting data from different data sources in the form of numbers, statistics and overall activity of any organization, with different deep learning and analytics tools, which is then processed using data visualization software and presented in the form of graphical charts, figures, and bars.

In today technology world, data are reproduced in incredible rate and amount. Visual Analytics helps the world to make the vast and complex amount of data useful and readable. Visual Analytics is the process to collect and store the data at a faster rate than analyze the data and make it helpful.

As human brain process visual content better than it processes plain text. So using advanced visual interfaces, humans may directly interact with the data analysis capabilities of today’s computers and allow them to make well-informed decisions in complex situations.

It allows you to create beautiful, interactive dashboards or reports that are immediately available on the web or a mobile device. The tool has a Data Explorer that makes it easy for the novice analyst to create forecasts, decision trees, or other fancy statistical methods.

#blogs #data visualization #data visualization tools #visual analytics #visualizing ml models

Visual Perception

Why do we visualize data?

It helps us to comprehend _huge _amounts of data by compressing it into a simple, easy to understand visualization. It helps us to find hidden patterns or see underlying problems in the data itself which might not have been obvious without a good chart.

Our brain is specialized to perceive the physical world around us as efficiently as possible. Evidence also suggests that we all develop the same visual systems, regardless of our environment or culture. This suggests that the development of the visual system isn’t solely based on our environment but is the result of millions of years of evolution. Which would contradict the tabula rasa theory (Ware 2021 ). Sorry John Locke. Our visual system splits tasks and thus has specialized regions that are responsible for segmentation (early rapid-processing), edge orientation detection, or color and light perception. We are able to extract features and find patterns with ease.

It is interesting that on a higher level of visual perception (visual cognition), our brains are able to highlight colors and shapes to focus on certain aspects. If we search for red-colored highways on a road map, we can use our visual cognition to highlight the red roads and put the other colors in the background. (Ware 2021)

#data-visualization #gestalt-principles #visualization #data-science #visual-variables