Sir Your tutorials are really amazing and I really learned many many things from you as a Computer engineering student. I wish I could become like you someday: A Good Person who has lots of KNOWLEDGE. Thank you so much...
I had understanding issues in bulk processing of images and while searching your channel, got into this........Just to ensure, I learn it first time exactly the way you taught, I replicated the same kind of folders, images, and including Tif Stack image and it took me around 2 hours from start to end......Have a hang of it, but must practice more times to gain speed in this.......At 21.40, you mentioned it trivial.......But nothing is trivial in these Ajarn........My humble pranams as always......
Hi Sreeni, thank you so much for these great videos. I have 100 images taken by a microscope, and want to take Fourier transformation for them and apply a numerical integral over the difference values over time, I would be appreciated if you could help me with that. I tried so hard, but cannot figure it out.
Thanks! Great help filling in the missing parts from other online courses. I was working on this part. A small correction, “pwd” is “print working directory,” copied from a Unix bash command.
Hi Sreeni sir, I have learnt a lot from your videos and I like you to teach us the tricks like this which will be helpful for the beginners (like me). I have involved recently in cancer detection (based on images) using image segmentation (But I am from other speech-related domain) and I believe masks creation and collecting images will be challenging for the tasks. 1.Can you share any tools that will be helpful for mask creations (like Apeer) 2. Experimenting purpose datasets That will be helpful as a beginner. Thanks
@@DigitalSreeni Hi Sreeni sir, I have said that I am from the speech domain. Currently, Working on cancer detection related tasks and I am a beginner in image segmentation. I have requested inputs on the above image segmentation task, not on the speech task
Great tutorial indeed thank you for the effort. Please is there any other tutorial where you talked about multiple images registration? especially geospatial images. Secondly, is this process under deep learning
Great tutorial. Thank you so much. How can we handle categorical features in multi variate time series scenario? Label encoding categorical column and then scaling it along with other numerical features or OHE ? If number of categories are more than 30 or so, OHE would increase size of data frame.
Thank you for your explanation, it's really helpful to do my dissertation. One thing that I am confused about is how can I save the processed images back to original name + str()? Ive tried to find your video about glob as you mention but I cant find it anywhere. Thank you once again!
brilliant job! I have learned a lot from the videos. I am doing the vesicle fluctuation analysis by python. need to deal with thousands of images at one moment. for this time, I added the Sobel method into each image process. I got the two lines for each circle edge, which should be from the bright one and the dark one (phase contrast microscopy). how could I remove one of the two lines....many thanks.
Hi, awesome tutorial! Thanks. I have one question though. What if I have multiple RGB images as one input, and how should I combine / overlap them correctly to train a convolutional neural network? Ex: The shape of one image is (24,24,3). If I combine four of them, the shape would be (24,24,12)?
I believe it needs an additional dimension. The shape of one array with only one 3 channel image would be [1,24,24,3] and when you have four images you would have [4,24,24,3]
Thanks so much for the awesome tutorial! I just had one question though, I am using colored images as my input, but then when I finish the output is grayscale. I saw that you said to add 1 more dimension in the array for the color, but how does that work exactly? I could not get it to work even after changing images_llist[image,:,:,:]
I covered these topics in this video but if you want separate videos, here are the links to the ones I created under my work account: ruclips.net/video/Z90KEqJoC3w/видео.html ruclips.net/video/j6GNtqrwcNE/видео.html
You need deep learning to get robust results. Here is a great paper on this topic. I will try to create a video on this in a few weeks. arxiv.org/pdf/2003.02261.pdf
Thanks a lot for this video Sir. I am very new at image processing. I have a question I have, say 100 images, and the camera position is fixed. The images are of soil moving very slightly in a box. Which one of your lectures would be best to carry out analysis such that I get a video of all images where soil particles are moving. Also if I know the grid size how can I determine the amount of movement let's say in mm. Thanks and regards
I am not sure if I have a specific video that addresses your problem. Normally you have algorithms for object tracking where you segment objects and track them when they move. I am not sure how to do the same for soil where everything may look the same and sometime things move.
@@DigitalSreeni actually i have placed small white particles big enough to be seen by naked eye. And they are placed in a grid of 5cm. So i thought you could help me tracking those particles through python. Nonetheless thanks for replying. And i highly appreciate your efforts. Regards
Thank you much for your tutorials! By the way, regarding the feature extraction stage, I recently saw one of your videos where you can use gabor features and export them to a csv file for one image. What should I do if I wanted to iterate, say, over 100 images and have those features exported to the same csv file?
I covered it in many of my videos. For example, the one where I explained feature based image segmentation uses Gabor to capture filtered images into a pandas dataframe. Youc an easily save the dataframe into a csv file. Here is one of those videos: ruclips.net/video/sD2xL36Xdu0/видео.html
Hi, can you please make a video on how to implement data augmentation for semantic segmentation in keras. As you mentioned last time that Keras ImageDataGenerator changes the values of the masks while transforming. Did you find any solution for that ? I am still looking for a solution of how to apply augmentation on the fly. Thanks
@@DigitalSreeni Thanks. BTW I applied Keras Image Generator on the fly for semantic segmentation binary classification. I could share with you. Please send me an email on umairsabir21@gmail.com
This is a wonderful video. Kudos to you. Please, could you make a video on how to find an average on different multiple similar images and form a template matching from the resulting image?
I am not sure what you are recommending. When you average multiple images you will either get garbage if images are different or a denoised image if images are all showing same field. Why is this related to template matching?
I am working on the hand region for recognition for the sign, I used canny edge detection to detect the edges in the image. So, my intention is to form a matching template from the acquired images (training dataset). That's the reason I was thinking of averaging images. What are better ways rather than using the area?
Hi can you do a tutorial to show how to create batched dataset for semantic segmentation using images stored locally rather than importing them to a numpy array. Because if we have 10k images we cant get them all to a numpy array.
Very informative video, is it possible to build 3D image by combining series of 2D images in python? like making tomographic view with various slices of 2D images. I would appreciate if you can make video tutorial on it. Thanks
By stacking a bunch of 2D slices (in numpy) wouldn't it effectively become a 3D image? If you are talking about using individual images of 2D projections (e.g. from x-ray) to reconstruct 3D volume then you can definitely do it. For that you need to know the theory behind FDK or iterative reconstruction. By the way, if you have segmented 2D images you can stack them to combine 2D segmented objects into 3D objects. We have a module on APEER exactly to do this task. www.apeer.com/app/modules/Measure-3D-Objects-Parameters/c0402c81-af4a-4143-88dd-fa40717df8f1
Really awesome Sir, I'm learning a lot by watching this channels videos day by day. Sir I need a suggestion on one thing. If I want to work on such Research topics and need to Draw or design nice quality picture or architecture like some research articles, what Software or Tool I should use to draw those. (Except PowerPoint is there any other options that could be helpful for newbie students to work on it. Thank you 😊
People normally use Microsoft Visio for drawing. A quick google search for free alternatives gave me this page: www.maketecheasier.com/5-best-free-alternatives-to-microsoft-visio/
Thank you very much for your videos. I am a neuroscientist who has benefit a lot from your tutorials. I would like to respectfully make a request. I have always imagined being able to subtract the background from my microscopy images, through the patterns of a negative control image, i.e. an image that shows me all the autofluorescence and noise, but not the cells of interest. In other words, to tell the computer: "This is how a typical image without cells will look like, now, subtract that from my images containing cells". It will be possible? Thanks
You seem to be referring to subtracting a background image from another image, if so this is a simple task of just subtracting 2 images. If you are referring to background and other features as features that are not of interest compared to cells, you can train a binary classifier with your cells as objects and everything else as background.
Hello SIr! Thanks for this great video! I actually tried this on my ultrasound images, the code is working but it's not saving the files into the specified path I wrote in the script. Would you mind checking this? #Gaussian Blur import os import cv2 from skimage.filters import gaussian from skimage import img_as_ubyte i = 1 for root, dirs, files in os.walk("Practicum/CROPPED/NORMAL"): for name in files: print(os.path.join(root,name)) path = os.path.join(root,name) norm_img = cv2.imread(path, 0) blur_norm_img = img_as_ubyte(gaussian(norm_img, sigma = 5, mode = 'constant', cval = 0.0)) cv2.imwrite("Practicum/GAUSSIAN BLUR/NORMAL/name"+ str[i] + '.jpg', blur_norm_img) i += 1
Thank you very much Sreeni !! I just made my very first loop to process 370 images. I really appreciate your effort for making this video
Sir Your tutorials are really amazing and I really learned many many things from you as a Computer engineering student. I wish I could become like you someday: A Good Person who has lots of KNOWLEDGE.
Thank you so much...
I don't know how to thank you enough Sir! Your videos are like answer to all my confusions.
Thank you so much for all the great work you are doing.
You are very welcome Rowzat. Please keep watching.
Thank you, respect from a Chinese student! your work is very outstanding。
Thank you very much!
I had understanding issues in bulk processing of images and while searching your channel, got into this........Just to ensure, I learn it first time exactly the way you taught, I replicated the same kind of folders, images, and including Tif Stack image and it took me around 2 hours from start to end......Have a hang of it, but must practice more times to gain speed in this.......At 21.40, you mentioned it trivial.......But nothing is trivial in these Ajarn........My humble pranams as always......
Happy to see you Back Teacher. You are my Inspiration
Thank you! 😃
Hi Sreeni,, I am a beginner to py prog. This video helped me a lot . Thanks for your valuable videos.........
Great great tutorial, i truly appreciate your work. Just started with microscopy work and it has proven to be very useful
Awesome video. A lot of basic things explained in detail. Much to learn from it. Thank you.
Glad you enjoyed it!
Amazing again Exactly what I needed!
Hi Sreeni, thank you so much for these great videos. I have 100 images taken by a microscope, and want to take Fourier transformation for them and apply a numerical integral over the difference values over time, I would be appreciated if you could help me with that. I tried so hard, but cannot figure it out.
Thanks! Great help filling in the missing parts from other online courses. I was working on this part. A small correction, “pwd” is “print working directory,” copied from a Unix bash command.
I'm not sure what I said about 'pwd' but you are correct, the command 'pwd' will print the present working directory on the screen. Thanks.
Hi Sreeni sir,
I have learnt a lot from your videos and I like you to teach us the tricks like this which will be helpful for the beginners (like me). I have involved recently in cancer detection (based on images) using image segmentation (But I am from other speech-related domain) and I believe masks creation and collecting images will be challenging for the tasks.
1.Can you share any tools that will be helpful for mask creations (like Apeer)
2. Experimenting purpose datasets
That will be helpful as a beginner.
Thanks
I am sorry, I do not have much experience nor do I know of any resources on speech related deep learning.
@@DigitalSreeni Hi Sreeni sir, I have said that I am from the speech domain. Currently, Working on cancer detection related tasks and I am a beginner in image segmentation. I have requested inputs on the above image segmentation task, not on the speech task
hello sir how can i copy the same file name from the original to the modified image?
thank you!
Thank you !! Love your videos they are my favorite ❤
Great tutorial indeed thank you for the effort. Please is there any other tutorial where you talked about multiple images registration? especially geospatial images. Secondly, is this process under deep learning
Great tutorial. Thank you so much. How can we handle categorical features in multi variate time series scenario? Label encoding categorical column and then scaling it along with other numerical features or OHE ? If number of categories are more than 30 or so, OHE would increase size of data frame.
Thank you for your explanation, it's really helpful to do my dissertation. One thing that I am confused about is how can I save the processed images back to original name + str()? Ive tried to find your video about glob as you mention but I cant find it anywhere. Thank you once again!
brilliant job! I have learned a lot from the videos. I am doing the vesicle fluctuation analysis by python. need to deal with thousands of images at one moment. for this time, I added the Sobel method into each image process. I got the two lines for each circle edge, which should be from the bright one and the dark one (phase contrast microscopy). how could I remove one of the two lines....many thanks.
Thank you very Much! this is just exactly what I was looking for! Much appreciatet and perfect for a beginner like me :)
sir can you suggest me how i can stitch multiple images as single image.
Thank you so much for this tutorial. This helped a lot!
Glad it helped!
Hi, awesome tutorial! Thanks.
I have one question though. What if I have multiple RGB images as one input, and how should I combine / overlap them correctly to train a convolutional neural network? Ex: The shape of one image is (24,24,3). If I combine four of them, the shape would be (24,24,12)?
I believe it needs an additional dimension. The shape of one array with only one 3 channel image would be [1,24,24,3] and when you have four images you would have [4,24,24,3]
I love all you videos❤️ .. could you please make video on times series image prediction/forecasting with lstm?
Never tried but good suggestion. I will add to the list.
How to add another dimension for color?
Thanks so much for the awesome tutorial! I just had one question though, I am using colored images as my input, but then when I finish the output is grayscale. I saw that you said to add 1 more dimension in the array for the color, but how does that work exactly? I could not get it to work even after changing images_llist[image,:,:,:]
Could you please give me the link to the other two videos on GLOB and OS? I cannot find them among your videos.
I covered these topics in this video but if you want separate videos, here are the links to the ones I created under my work account:
ruclips.net/video/Z90KEqJoC3w/видео.html
ruclips.net/video/j6GNtqrwcNE/видео.html
how can i work on diabetic retinopathy images detection with segmentation techniques ?
You need deep learning to get robust results. Here is a great paper on this topic. I will try to create a video on this in a few weeks. arxiv.org/pdf/2003.02261.pdf
Thanks a lot for this video Sir.
I am very new at image processing. I have a question
I have, say 100 images, and the camera position is fixed. The images are of soil moving very slightly in a box. Which one of your lectures would be best to carry out analysis such that I get a video of all images where soil particles are moving. Also if I know the grid size how can I determine the amount of movement let's say in mm. Thanks and regards
I am not sure if I have a specific video that addresses your problem. Normally you have algorithms for object tracking where you segment objects and track them when they move. I am not sure how to do the same for soil where everything may look the same and sometime things move.
@@DigitalSreeni actually i have placed small white particles big enough to be seen by naked eye. And they are placed in a grid of 5cm. So i thought you could help me tracking those particles through python. Nonetheless thanks for replying. And i highly appreciate your efforts. Regards
Thank you much for your tutorials! By the way, regarding the feature extraction stage, I recently saw one of your videos where you can use gabor features and export them to a csv file for one image. What should I do if I wanted to iterate, say, over 100 images and have those features exported to the same csv file?
I covered it in many of my videos. For example, the one where I explained feature based image segmentation uses Gabor to capture filtered images into a pandas dataframe. Youc an easily save the dataframe into a csv file. Here is one of those videos: ruclips.net/video/sD2xL36Xdu0/видео.html
Hi, can you please make a video on how to implement data augmentation for semantic segmentation in keras. As you mentioned last time that Keras ImageDataGenerator changes the values of the masks while transforming. Did you find any solution for that ?
I am still looking for a solution of how to apply augmentation on the fly. Thanks
I will record a video but please try albumentations library. Their documentation is very helpful.
@@DigitalSreeni Thanks. BTW I applied Keras Image Generator on the fly for semantic segmentation binary classification. I could share with you. Please send me an email on umairsabir21@gmail.com
This is a wonderful video. Kudos to you. Please, could you make a video on how to find an average on different multiple similar images and form a template matching from the resulting image?
I am not sure what you are recommending. When you average multiple images you will either get garbage if images are different or a denoised image if images are all showing same field. Why is this related to template matching?
I am working on the hand region for recognition for the sign, I used canny edge detection to detect the edges in the image. So, my intention is to form a matching template from the acquired images (training dataset). That's the reason I was thinking of averaging images. What are better ways rather than using the area?
Hi can you do a tutorial to show how to create batched dataset for semantic segmentation using images stored locally rather than importing them to a numpy array. Because if we have 10k images we cant get them all to a numpy array.
This will be the topic of my upcoming video which will be numbered 222. Please stay tuned and hopefully you subscribed to the channel.
@@DigitalSreeni Wow can't wait to watch it. Thank you for sharing your work & knowledge...
Hi Sreeni! Good to see you back again. Can you make a tutorial on how to use PCA to speed up traditional ML?
nvm, i realized you already did a few videos on that... Thanks so much!!
Very informative video, is it possible to build 3D image by combining series of 2D images in python? like making tomographic view with various slices of 2D images. I would appreciate if you can make video tutorial on it. Thanks
By stacking a bunch of 2D slices (in numpy) wouldn't it effectively become a 3D image? If you are talking about using individual images of 2D projections (e.g. from x-ray) to reconstruct 3D volume then you can definitely do it. For that you need to know the theory behind FDK or iterative reconstruction.
By the way, if you have segmented 2D images you can stack them to combine 2D segmented objects into 3D objects. We have a module on APEER exactly to do this task.
www.apeer.com/app/modules/Measure-3D-Objects-Parameters/c0402c81-af4a-4143-88dd-fa40717df8f1
@@DigitalSreeni Thank you. It looks like stacking 2D slices in numpy or the module in APEER should work.
Hello Sir,
Hope that you would be fine. I want to know that how can i make a 3d panorama image by using multiple 2d images.
Thanks In Advance
I do not know how to create 3D panorama image using 2D images.
Really awesome Sir, I'm learning a lot by watching this channels videos day by day.
Sir I need a suggestion on one thing. If I want to work on such Research topics and need to Draw or design nice quality picture or architecture like some research articles, what Software or Tool I should use to draw those. (Except PowerPoint is there any other options that could be helpful for newbie students to work on it.
Thank you 😊
People normally use Microsoft Visio for drawing. A quick google search for free alternatives gave me this page: www.maketecheasier.com/5-best-free-alternatives-to-microsoft-visio/
@@DigitalSreeni thank you so much sir. Much helpful advice from you. Very much grateful 😊
hi sreeni.. i have some doubts
hi sreeni
Thank you very much for your videos. I am a neuroscientist who has benefit a lot from your tutorials. I would like to respectfully make a request. I have always imagined being able to subtract the background from my microscopy images, through the patterns of a negative control image, i.e. an image that shows me all the autofluorescence and noise, but not the cells of interest. In other words, to tell the computer: "This is how a typical image without cells will look like, now, subtract that from my images containing cells". It will be possible? Thanks
You seem to be referring to subtracting a background image from another image, if so this is a simple task of just subtracting 2 images. If you are referring to background and other features as features that are not of interest compared to cells, you can train a binary classifier with your cells as objects and everything else as background.
Thank you very much
You are welcome
Thank you so much.
Always welcome
Awesome thanks
No problem!
Hello SIr! Thanks for this great video! I actually tried this on my ultrasound images, the code is working but it's not saving the files into the specified path I wrote in the script. Would you mind checking this?
#Gaussian Blur
import os
import cv2
from skimage.filters import gaussian
from skimage import img_as_ubyte
i = 1
for root, dirs, files in os.walk("Practicum/CROPPED/NORMAL"):
for name in files:
print(os.path.join(root,name))
path = os.path.join(root,name)
norm_img = cv2.imread(path, 0)
blur_norm_img = img_as_ubyte(gaussian(norm_img, sigma = 5, mode = 'constant', cval = 0.0))
cv2.imwrite("Practicum/GAUSSIAN BLUR/NORMAL/name"+ str[i] + '.jpg', blur_norm_img)
i += 1