Thanks Sreeni for your series on BraTS dataset. I have been looking forward to it, esp on your next section on the data generators. You are right as the tensorflow data generators only handles RGB. Some of the modifications to the tensorflow Class is too complicated. Hope to see your next installment real soon.
Thanks, for the videos, really enjoying the semantic segmentation of different parts of the body, could you also do one with bone marrow fibrosis as fibrosis seems hard to do when getting labelled data?
Please do share more videos how to do semantic segmentation with BraTS dataset, looking forward to more videos on this topic. Please do make videos from starting to end with this dataset.
My question isn't related to this video and might be a dumb question, sir. I'd like to know if it is possible to ensemble( for the lack of a better word) pretrained CNNs(like VGGnet, Inceptionv3) and combine the best features recognised easily by each pretrained CNN and then fed into a classifier for image classification? Or is that a bit overkill?
6:05 In the paper titled "The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS)" Pg 8, they describe the annotation in BraTS 2012/13 datasets. It seems that they are using 5 labels in the mask there: Normal or no tissue (label 0) Tumor Core (label 1, manually annotated by experts using T2w) Edema (label 2, whole tumor visible in FLAIR - Necrotic non-enhancing Solid Core NCR visible in T2w) Enhancing Tumour Core (label 4, manually annotaed by experts using Gandolinium-enhanced T1w aka T1Gd or T1ce) Cystic Necrotic components of the Core (sub-component of Enhancing Tumor Core as seen from T1Gd, but they have not explained the criteria) It is possible that Label 3 was that in older BraTS datasets, and no longer annotated in newer BraTS datasets (starting 2017). Just some guesswork.
Thank you professor for making so many useful videos. I am a student who is working on medical image segmentation as my final year project. These videos really make my life easier. Could you please consider doing a bunch of videos about LIDC-IDRI dataset, cause as a beginner in this field, I am really confused about using it. I would really appreciate for this.
Hey sreeni, I am working on segmentation models and installed 0.2.1 version. But when I try to run losses, it says segmentation_models module has no DiceLoss attribute. Could you help me with this?
Whenever you have such questions, you need to look at the documentation. According to their code for losses, they do have DiceLoss. So please check your code to see if you are calling it properly. github.com/qubvel/segmentation_models/blob/master/segmentation_models/losses.py
brother you are great i am from india your videos are the only way to learn from zero to hero i am doing my project on aerial image classification please make one video for google earth imagery classification thank you
hi, Mr. Sreeni, how do define a custom mask(manually) for an image, for example, I have license-plate images and I would like to create masks for it to segment different components of them. thanks for considering, behnoud
I use www.apeer.com to load my images and paint the pixels I'd like to assign to a specific class. PS: APEER is developed my team at work and it is free.
![Quando Rondo - Grow Up [Official Music Video]](http://i.ytimg.com/vi/8zFnCg3BO4Q/mqdefault.jpg)
Thank you prof, this content is very important to me and it comes on time, Great channel for learners.
You are very welcome
Thanks Sreeni for your series on BraTS dataset. I have been looking forward to it, esp on your next section on the data generators. You are right as the tensorflow data generators only handles RGB. Some of the modifications to the tensorflow Class is too complicated. Hope to see your next installment real soon.
Yes, soon
Thanks, for the videos, really enjoying the semantic segmentation of different parts of the body, could you also do one with bone marrow fibrosis as fibrosis seems hard to do when getting labelled data?
Please do share more videos how to do semantic segmentation with BraTS dataset, looking forward to more videos on this topic.
Please do make videos from starting to end with this dataset.
More on the way.
@@DigitalSreeni Your videos help students like me who are eager to learn new things and I can't thank you enough for your extraordinary work.
I tried to apply this dataset to segment by using ResNet101 but not succeed.
Thank you very much, Professor Sreeni. All of your tutorials are very helpful and excellent!
My question isn't related to this video and might be a dumb question, sir.
I'd like to know if it is possible to ensemble( for the lack of a better word) pretrained CNNs(like VGGnet, Inceptionv3) and combine the best features recognised easily by each pretrained CNN and then fed into a classifier for image classification?
Or is that a bit overkill?
My best youtube channel learning a lot
Thanks, keep learning.
Thank you so much prof for these quality contents. Looking forward to learn so many new things!
6:05 In the paper titled "The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS)" Pg 8, they describe the annotation in BraTS 2012/13 datasets. It seems that they are using 5 labels in the mask there:
Normal or no tissue (label 0)
Tumor Core (label 1, manually annotated by experts using T2w)
Edema (label 2, whole tumor visible in FLAIR - Necrotic non-enhancing Solid Core NCR visible in T2w)
Enhancing Tumour Core (label 4, manually annotaed by experts using Gandolinium-enhanced T1w aka T1Gd or T1ce)
Cystic Necrotic components of the Core (sub-component of Enhancing Tumor Core as seen from T1Gd, but they have not explained the criteria)
It is possible that Label 3 was that in older BraTS datasets, and no longer annotated in newer BraTS datasets (starting 2017). Just some guesswork.
Thank you for this video. Have one doubt. After cropping 128*128*128. what does last 128 indicates? we have only three channels why 128 here.
Thanks prof for these videos, I am doing this challenge and I hope you can continue this series as soon as possible.
That's the plan!
Please do this dataset with survival time. Thank you
Interesting Topic! I liked the way you teach things! Waiting for next video
Hi Sreeni, Once again great video series.
Requesting you for a MultiModal classification with Image and tabular data as input.
hello sir, you are doing great...I am looking forward to it.
You are indeed genius...keep doing the great work
Great work... please create similar videos for UCSF_BrainMetastases dataset
I am working in medical imaging. What are your system specifications. Do you recommend a particular laptop or a customized setup for such processing?
FINALLY! THANK YOU SREENI!
You are welcome. Keep watching.
hey teacher u r amazing. please make some stuff on lung cancer prediction
Thank you professor for making so many useful videos. I am a student who is working on medical image segmentation as my final year project. These videos really make my life easier. Could you please consider doing a bunch of videos about LIDC-IDRI dataset, cause as a beginner in this field, I am really confused about using it. I would really appreciate for this.
Hi brother. I have chosen brain tumor detection as my fyp. I need some help.
Thanks sir for the informative video. Sir, can you please make some videos related to Vision Transformers?
Train the modal of cmct in this dataset with 240 subjects where 200 for training and 40 for testing.plz it's a request
Make a Video by using same dataset to segment by ResNet101 model….
Hey sreeni, I am working on segmentation models and installed 0.2.1 version. But when I try to run losses, it says segmentation_models module has no DiceLoss attribute. Could you help me with this?
Whenever you have such questions, you need to look at the documentation. According to their code for losses, they do have DiceLoss. So please check your code to see if you are calling it properly.
github.com/qubvel/segmentation_models/blob/master/segmentation_models/losses.py
You are a rockstar. Thank you.
what should be the minimun system requirement to do model training?
brother you are great i am from india your videos are the only way to learn from zero to hero i am doing my project on aerial image classification please make one video for google earth imagery classification thank you
I am not getting why to rename that ?
Thanks for the explanation bro
Just love your contains sir
Thanks a ton
Thank you very much
Is this code compatible with nifti file format?
Hai sir i'm getting difficulty while trying to fit the model , how much hardware it might require to train the model?
At least 16GB GPU. May be you get away by using the free version of Colab.
Good job keep it up
Thanks
hi, Mr. Sreeni, how do define a custom mask(manually) for an image, for example, I have license-plate images and I would like to create masks for it to segment different components of them.
thanks for considering,
behnoud
I use www.apeer.com to load my images and paint the pixels I'd like to assign to a specific class.
PS: APEER is developed my team at work and it is free.
@@DigitalSreeni Thanks for answering Mr.Sreeni.