DevoWorm Tutorials, 29 Aug 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Tutorial for DevoLearn

Description

Ujjwal Singh provides a guided tour of the DevoLearn platform

A

Hi everyone- uh I am singh- and I am today here to give you a brief introduction about our newly formed python package that is duolearn.

A

So, as you can see like, we have created a small github page for the table, learn where we have different repositories for different purposes, like education theory, building contribution guidelines which contains like how to onboard for a new user or media like c dot elegance table loan is a web app for cd and segmentation procedure, data science demo in development library.

A

So basically, today we are going to be focused about table one library, so let's dive a little bit into it.

A

So, as you can see like this is a table and library, it has a very detailed uh readme, so the for the new contributors or can just come and like see things uh who have a little bit of experience in coding. They can easily get all the things done here. So these are the maintenance and contributors. You can contact us anytimes uh and we are ready to help you so my my name is uh moving on. So basically, this tutorial is mean for the people who are not very much familiar with the computer science.

A

So I'm going to just give you a brief uh walkthrough like how to use this library. So as the load suggests itself, it is a library which is focused to accelerate the data-driven developmental biology research of computational learning models. So right now, this library is basically focused on c dot elegance segmentation procedure. So we are going to talk about it most of the time. So let me just quickly open the column notebook, so you can use any notebook like collab. Notebook is a online free tool which is very useful.

A

Sometimes so you can use that other than that or you can use jupyter notebooks if you have installed it, and you can also use simple python scripts to run this. But then the project is going to be a little bit different because, like we are going to install library within the notebook, then you have to install that little terminals and command line whatever system you are using.

A

So let's quickly name over it as intro to.

B

They will earn.

A

So first thing that we need to do before anything doing it is to install the package.

A

So installing pack is very much easy in this case because, like it is a pip package that you can readily install using pip install table learn so, let's see oh, I have this button is to run so this basically runs the block in which you have written the code. So, as you can see, it is right now connecting once it is connected. It will run this blog and search for the library or they will learn, and we have hosted it on the pi pi packet.

A

So it is like downloading it and like installing all the requirements that are needed. So, as you can see, it is successfully installed once successfully installed. You can like go to the next.

A

So basically, our tutorial this time is going to be focused on, uh like the explorations that we are going to do with the development like the objective for this exercise or for this introduction is to like we try to visualize the distances between the two centroids of each cell in cedar, elegant embryo during the embryogenesis process, and I try to make a 3d model of c dot elegance emra, using images from like different planes of the same embryo.

A

Let me first quickly install some libraries and dependencies that I need so I have imported os. I have imported matplotlib, so like these things are basically uh depends on what kind of work you want to do with it. So we also need pandas.

A

So like I am writing this like uh import pandas as pd and matplotlib as plt. So basically you can. These are the uh abbreviations that we give to different libraries. So these are not basically required, like you can give a pd to pap and whatever you like, or you can simply use pandas, but for this like instead of typing pandas, again and again, you can use small applications.

A

So let me import cv2. We don't need to give give name a list because it already is very short name.

B

Then let us import numpy.

B

B

Oh stop before I don't miss my thing.

B

A

A

So, as you can see like most of these tools that I am importing is basically to visualize things, ah so let us import these things, so you can also run like. Oh, we get error. So that's a typing mistake like these things can be avoided.

B

Oh, I think this should work.

A

So all the library pack and packages are instantly installed and these all things are done after that you just need to import like we are going to deal with the umbra segmentation, so the module or the model name in the library they will learn is basically named as.

A

A

So, as you can see, it has automatically completed it and after that you can store this in the model like.

B

So we have successfully saved the model.

A

In it is like downloading, it.

A

Moving on oh now, we need to load and extract images from like google drive so like I, uh we have saved some images on the google drive for the data you can uh take whatever like you want to use. Whatever you kind of I mean you want to use.

B

So let me just quickly copy paste for everything we need.

A

And after that, like you can insert your drive name here, then you can like uh import your training data that you have of the c dot elegance uh or whatever data you are using. You can just import and unzip it so to unzip. We have a command like uh control, unzip.

A

So if it is in a zip format, you can quickly unzip it and replace this. Whatever data set name with the dataset name, that you have.

B

A

After that, like after you have successfully unzipped your data set, you can check whether a data set is working or not like you can import your data set in pandas, so like bf, is the data frame. Basically, we are creating a data frame from pandas, so partners we have abbreviated as pd, so pandas dot read csv.

A

Whatever your csv file name is, you can just replace whatever file with the csv file name that you have, and you can even see it like if the data set is loaded or not. So this will show you like if successfully loaded, it will show you some entries of your data set, so you can know like whether the set is loaded or not.

A

Then like we move forward uh to like selecting the images of the single embryo for the experimentation, so we should like try to extract the centroids of the single embryo cells using uh undergo embryogenesis process.

A

So let me now just uh quickly after showing you, the basic setup quickly drag the jupyter notebook that we are working with. So let me just quickly show you yeah. So, basically, as you can see, there.

A

So, as you can see there, we have basically different selecting for the experimentation so, like you can import your image path here and after image part, you can replace the images and after that, you can give the model predictions about sg and centroids. So after that we have set the centroid mode. True, we have to visualize the centroid which are determined by the model and they will learn so we have basically like, first of all, shape whatever uh image path and hd.

A

If we have in the numpy array, and then we have created some distance metrics, as you can see, you have put ordered and centroids here the point of centroids here and after that you can see like uh from the plot show. We can easily visualize like what are the centroids and how they are connected. So it's basically a very basic graph in which we have connected every centroid to every centroid, but you can definitely change these parameters using the uh foreign range functions for above for two for loops. You can basically change that.

A

Then there is a basic distance matrix that you can use to visualize your outcomes.

A

Then so like this is the little experiment that you can do with the table. Learn certain can do a lot more it just to get you started and like how to install and all those things. So you can always visit like these notebooks are created by like one of our team member and he has hosted these notebooks in the github library.

A

So you can definitely go there and see all those things like how they have completed these things like creating animations and all so like.

A

These are the detailed things that uh most of the people like who are very much interested, can go in the library and like want to know how the things work can go through for others like they can simply do all these things with uh like by just visiting the readme.html, who are not a computer scientist, but they are biologists, but they want to use this thing, so they can just go to the redmi. We have tried to make as detailed me as we can.

A

As you can see, all the things are listed here, like first of all, segmenting a c dot in elegance embryo. Not a big deal, I have shown you already like how to set up a devolent library on your jupiter column, notebook for terminal. You have to do the same thing just instead of uh exclamation mark you have to just type this part in your terminal where you want to install it and rest will be taken care of by pipe and package itself.

A

So, as you can see like generating synthetic images, you can just import the model and you can create visualizations and graphs, so predict the population of cell within the seed or delegate. So this thing you just need to follow these steps, so this is the basic uh tutorial that we have to give to community so that we can get started. If you have any questions that you need, you can contact us any of us, maintainer, myokutube or dr bradley. We are very happy to answer you.

A

You can also join our slack workspace if you want, if you want to contribute to this course, so I think that's all from my side right now.

A

Thank you for joining.

B

B