20 - Introduction to image processing using scikit-image in Python

Рет қаралды 71,504

Күн бұрын

Scikit-image is a Python library dedicated towards image processing. This video explains a few useful functions from the scikit-image library including, resize, reshape, edge detectors and segmentation process for a microscopy based assay (wound healing or scratch assay).
The code from this video is available at: github.com/bnsreenu/python_fo...

Пікірлер: 81

@bluecubeist 4 ай бұрын

I love this channel. It is adding real value to my life by increasing my skills and knowledge, and that makes me happy. Thank you from the bottom of my hear sir.

@texasfossilguy Жыл бұрын

Exceptional video. I woupd love an extended one which would go through that entire library. That is so useful.

@GeorgeC1andonly 3 жыл бұрын

Thanks again Sreeni - I am planning to get through all of your videos - you have made SO many this year!!!!

@DigitalSreeni 3 жыл бұрын

Normally I travel around the world for my job but due to lockdown I am home on the weekends and got some time to make these videos. I am glad you are learning.

@TeotoniodeCarvalho 4 жыл бұрын

This is absolutely amazing! Very good work!

@nathaliakim4778 4 жыл бұрын

These videos are very good, thank you for making this content!

@AlessandroBottoni 4 жыл бұрын

Great video, congrats! Very clear, very useful. Many, many thanks for the time you devoted to this task.

@DigitalSreeni 4 жыл бұрын

Glad it was helpful!

@f0wh516 3 жыл бұрын

Great videoseries, and it's useful way beyond microscoptics.

@felip6180 4 жыл бұрын

Amazing, specially with the problem solution you came up!

@zeyadramadan8759 2 жыл бұрын

This video is really to the point. I was trying to know how to analyze images in general and you made it to the point indeed. Thanks a lot.

@DigitalSreeni 2 жыл бұрын

Great to hear!

@yiding951 3 жыл бұрын

great tutorials! And thank people like you to share your knowledge!

@DigitalSreeni 3 жыл бұрын

What's point of gaining knowledge if you can't share it and what's the point of earning money if you can't support charities?

@muslumyildiz5694 Жыл бұрын

Informative and fantastic video.. Thanks..

@garfield111garfield Жыл бұрын

Hi. this is amazing... gives a lot of insight into image processing!

@arienugroho430 2 жыл бұрын

Thank's, For Your Tutorial's, It's Help Me A lot

@daljeetsinghranawat6359 9 ай бұрын

Very helpful.....Thank you........I am follwing this series of yours from the start....i want to do a project on image segmentation for M.Tech.

@alaasaid6404 3 жыл бұрын

This is amazing! Thank you so much

@DigitalSreeni 3 жыл бұрын

No problem 😊

@postnetworkacademy Жыл бұрын

Nice explanation

@koraycill 3 жыл бұрын

Explanatory and good video

@delos6329 3 жыл бұрын

Good video, thank you sir!

@DigitalSreeni 3 жыл бұрын

You are welcome!

@zeeshanahmed3997 4 жыл бұрын

Hello!! I have a question that "when I read my image as RGB i.e. img = imread("C:/Users/hp/Desktop/22.jpg") not in greyscale i.e. img = imread("C:/Users/hp/Desktop/22.jpg", as_gray=True), error occured (`block_size` must have the same length as `image.shape`.). How can I manage dimensions? Thank you

@DigitalSreeni 4 жыл бұрын

It doesn’t sound like an issue with reading the image but rather with some other operation you’re trying to perform on the image. Please run one line at a time to identify where exactly the issue shows up. For example, this could happen if you try to resize an image into a size that is not possible mathematically, like the input file dimensions not divisible by a scaling factor that you’re trying to use.

@rajputShailesh007 2 жыл бұрын

any tutorial to demonstrate online augmentation! Is it possible to manipulate images using without loading in Local storage? like just calling the dataset using library and processing the image

@mahhhhh2599 3 жыл бұрын

Hi sir, i already watch most of your videos. But i would like to know, how to do classification based on data that we manually extract the feature ? and one more thing, does gabor filter is a type of filter that extract features?

@DigitalSreeni 3 жыл бұрын

I already covered features and how to do segmentation using manually extracted features on my channel. Please watch videos 50 - 67. Yes, Gabor is a convolutional filter that can generate many features which can be used in machine learning.

@losamfr 3 жыл бұрын

Very interesting video. At the end, how can you count the number of cells using a certain threshold?

@isawatanabe5157 2 жыл бұрын

The binary operation classifies each cell as 1 or 0 depending on whether it is above or below the threshold. Just count either the ones or the zeros depending on whether you want to count cells above or below the threshold.

@saptarsheesarkar1657 4 жыл бұрын

why we are doing threshold on the image after the entropy rather than applying it on the original image ? what is the reason behind it? although you explained a bit but I didn't got it very well.

@DigitalSreeni 4 жыл бұрын

I guess I may not have explained it well then!!! If you could threshold the original image then why do you need tricks like applying entropy filter? As the original image is non-segmentable we are converting it into something that is segmentable, which is Entropy filtered image in this video.

@lucarubinetti2523 3 жыл бұрын

Hi Thanks for the videos. I have a question. I read the image with : "image = skimage.io.imread("image_path")". I'm trying to understand "skimage.transform.integral.integrate()" in order to calculate the integral sum image. My problem is how to select the coordinates : Top-left corner of block to be summed and the bottom-right corner of block to be summed inside the array image. Thanks a lot

@texasfossilguy Жыл бұрын

If its a numpy array of shape [3, 1000, 1000] you would use [3, [1:50], [1:50]] for the top left. If its a numpy array you can select it just like you select range values in any numpy array. If its an object then its probably similar. Try that code with and without the internal [ ], I am not sure which works.

@kimiaabdollahi23 Жыл бұрын

Hi... i tried to open images using scikit but ModuleNotFoundError: No module named 'skimage' appeared.... what should i do?

@treebeexyz 10 ай бұрын

seems that it changed names, im wondering what it is too

@ajaypatro1554 9 ай бұрын

what about syder using standalone i don't know much I am having troube time able to install any lib in it

@zeeshanpatel6358 4 жыл бұрын

This is a minor question, but at 17:19 you wrote deconvolved, _ = restoration.unsupervised_wiener(img, psf). Why is the underscore needed after deconvolved?

@DigitalSreeni 4 жыл бұрын

It is customary to use an underscore if you are assigning parameters to outputs of operations that yield multiple outputs but you don't care about some of them. In other words, the above operation of unsupervised_wiener outputs 2 values but we only care about the first one (deconvolved), so we assigned a name for deconvolved and for the second output we need to supply something, so we just gave an underscore.

@zeeshanpatel6358 4 жыл бұрын

Thank you for the answer

@manalihiremath2805 3 жыл бұрын

`block_size` must have the same length as `image.shape`. I am getting this error while rescaling,sizing and downscaling

@DigitalSreeni 3 жыл бұрын

I believe the block size you are using for resize or other operations need to be same size (length of array) as image dimensions. If your image dimensions are 3 (e.g. 3 x 3 x 4) then your block size must have a length of 3 (e.g. 4 x 1 x 3)

@justinlee1348 3 жыл бұрын

one question: I keep getting an error "module 'scipy' has no attribute '_lib' " and I think it's when I type 'from skimage.transform... from skimage.filters' . Why can I import skimage library fine by itself but get an error when I use 'skimage.transform ' etc? Maybe something wrong with my skimage.metrics? I tried to look online. I'm using python 3.7 on spyder btw

@DigitalSreeni 3 жыл бұрын

Python libraries constantly get updated. Therefore, you may find a function or module depreciated in some versions. Usually it displays a message about depreciation and suggests what to do. If not, please Google search for the problem and you'd find answer right away.

@mahmoudbenmami4131 4 жыл бұрын

10/10

@samarda3867 3 жыл бұрын

This is really just having some issues in that when I try to run the canny detection I get an error that states that my dtype cannot be converted to float. Is there any way to go about this error?

@samarda3867 3 жыл бұрын

Awesome!*

@DigitalSreeni 3 жыл бұрын

What is your dtype? Normally, you can convert any dtype to anything. Also, try loading images using skimage or opencv and see if that helps.

@mohitkumaryadav5351 4 жыл бұрын

requesting you to kindly keep the code at any repository for each video....that will be helpful for hand's on later

@DigitalSreeni 4 жыл бұрын

Yes, I've been intending to publish my code on Github but I need to find time to properly comment it first, I hate to share uncommented code. One of these days I'll find time to publish at least a few important videos where code can get a bit complex.

@monikasingh29 2 жыл бұрын

Sir, do you know any database which has microscopic images for leaves

@DigitalSreeni 2 жыл бұрын

Sorry, no I do not know of any such data sets. Please search for them on kaggle, you may have some luck.

@gedtoon6451 Жыл бұрын

Good video, but I don't like imports in the body of the code. The pep8 guidelines recommend all imports should be at the top of the file.

@GeorgeC1andonly 3 жыл бұрын

I tried this with some images I got from some microscopes (which were tiffs imported and were imported as type Uint16 ) and my deconvolved results were just 0 value images. When I took the original image and saved it as a jpg and opened it opened the image not as Uint16 but as float64 - then deconvolution worked - So - the solution was to convert the data by copying the array with something like copy=img.astype(np.float64)/65535 and to then use this new array to deconvolve- if anyone else has this issue - not an unimportant thing where users may be looking to load from 16bit tiff files as they are opened as Uint16 be default instead of converted to float64

@DigitalSreeni 3 жыл бұрын

When you do math to pixel values I recommend converting integers to float to not lose information due to rounding errors. Also, rescaling to values between 0 and 1 will help. These type of functions are built-in when you convert image from int to float using scikit image.

@JW-kl2wd 2 жыл бұрын

that is amazing, but i've tried this entropy filter on other pictures and i got this ValueError: ValueError: Image dimensions and neighborhood dimensionsdo not match! So is that normal? i just did not understand it. Thanks so much for this Video

@tevfikatrn 2 жыл бұрын

when you are reading image, you forgot to write as_gray= "gray"

@HarshaVardhan-es2fg 3 жыл бұрын

Can you send me link or reference for absolute beginners to get to know more about the deconvolution concept and point spread concept

@DigitalSreeni 3 жыл бұрын

I don't know of any references, I recommend searching for them online. A quick search gave me this document... spider.ipac.caltech.edu/staff/fmasci/home/astro_refs/Deconv_comparisons.pdf

@czeyveli1 2 күн бұрын

Could you share your default cell image?

@alessioandreoli2145 4 жыл бұрын

How to calculate the area of an object in the image?

@DigitalSreeni 4 жыл бұрын

Please watch my video on this topic... kzfaq.info/get/bejne/q5meepiQqtDKfWQ.html

@rinkalshah7303 4 жыл бұрын

My image I not showing in ipython console like yours...why?

@DigitalSreeni 4 жыл бұрын

I have no idea how to help, please share more details. If you’re using pyplot then try adding plt.show() at the end.

@ishanksachdeva369 4 жыл бұрын

ValueError: `block_size` must have the same length as `image.shape`.

@DigitalSreeni 4 жыл бұрын

Looks like you are trying to do some math on an image where the image size is not divisible by whatever number you are trying to use. In other words, please check the image shape (size) and see if you are properly dividing it into block size that makes sense.

@jaideeppanwar2730 3 жыл бұрын

sir I am getting a error :-ValueError: The parameter `image` must be a 2-dimensional array

@DigitalSreeni 3 жыл бұрын

The error is saying that the parameter you assigned as image is not a dimensional array. Please check if that is the case. May be you are using a color image and reading it as color so the dimensions are not just 2 and it is probably 3. If that is the case read image as gray; just add this to your imread... as_grey=True. For example: image = io.imread('my_image.jpg', as_grey=True)

@bhanunjaykumar1856 3 жыл бұрын

@@DigitalSreeni as_gray not as_grey

@eessaaabrahams9124 3 жыл бұрын

it says "No module named 'skimage' "

@sandalasanthisree6242 10 ай бұрын

I'm getting ValueError : Image dimensions and neighborhood dimensions do not match

@vt9848 2 жыл бұрын

Very good explanation with a lot of information. Small suggestion you can explain more clearly by drawing some images in paint or something else. Request: Can you please make some videos related to Hyperspectral images for feature extraction and selection

@DigitalSreeni 2 жыл бұрын

Hopefully you'll find that my later videos improved a bit.

@prashantmehta2832 4 жыл бұрын

Hi.. sir.. first i want to tell you thanks for this beautiful playlist of IM. But, i have a question.. A few months ago Chandrayaar-2 was lost on moon and a person found it on the moon.. my question is that... did he use image processing? and, if yes.. then can you tell me macroscopics image processing will help?(however, i am gonna learn and watch those all video bczz its really interesting and will help me some day ). and if no.. then plss tell me what should i do...

@DigitalSreeni 4 жыл бұрын

Not sure how they 'found' the lost Chandrayaan on the Moon, so I cannot comment. But, in general you need some sort of image analysis to identify features in an automated way.

@prashantmehta2832 4 жыл бұрын

@@DigitalSreeni Thanks sir.. For the information.. And right now I am at opencv.. I saw other OpenCV's videos and know OpenCV appropriately... But your videos are different which are more informative...

@samk4584 3 жыл бұрын

Hi is there a way we can contact you for a paid request ?

@DigitalSreeni 3 жыл бұрын

I’m sorry, I don’t do paid projects as I have a full time job that needs my full attention. You can definitely message me on LinkedIn or Twitter (@digitalsreeni). You can also send an email, address given as part of my profile on this channel. I can offer my advise, if I have knowledge in that area.