336 - Nuclei segmentation and analysis using Detectron2 & YOLOv8​

This video tutorial is an entire project spanning from data download to training object detection models to analysis and plotting. It covers the following key tasks, with downloadable code for every task:​
- Downloading data from Kaggle​
- Cleaning up the data​
- Converting masks to coco json and YOLOv8 annotations​
- Visualizing annotations​
- Training Detectron2 (Mask R-CNN) for object detection​
- Training YOLOv8 for object detection​
Code is available here: github.com/bnsreenu/python_fo...
Dataset downloaded from: www.kaggle.com/datasets/ipate...
Dataset description: arxiv.org/abs/2308.01760
Summary of the dataset:
The NuInsSeg dataset contains more than 30k manually segmented nuclei from 31 human and mouse organs and 665 image patches extracted from H&E-stained whole slide images. We also provide ambiguous area masks for the entire dataset to show in which areas manual semantic/instance segmentation were impossible.
Human organs:
cerebellum, cerebrum (brain), colon (rectum), epiglottis, jejunum, kidney, liver, lung, melanoma, muscle, oesophagus, palatine tonsil, pancreas, peritoneum, placenta, salivary gland, spleen, stomach (cardia), stomach (pylorus), testis, tongue, umbilical cord, and urinary bladder
Mouse organs:
cerebellum, cerebrum, colon, epiglottis, lung, melanoma, muscle, peritoneum, stomach (cardia), stomach (pylorus), testis, umbilical cord, and urinary bladder)

[@ImRiz1]

I've searching for good tutorials on Computer Vision since 2020 and found this channel only few days ago😒😒. I'm always late in the game. This channel is a Gem

[@user-od1pv8ib1o]

Thank you for your professionalism and clear presentation! I'll wait for further videos

[@DigitalSreeni]

Thank you very much :)

[@tahirak.7565]

U heard me without even saying anything. U really are a gem Sir.

[@DigitalSreeni]

Yeah, I can read minds :)

[@maxgadd]

nice work man. thanks

[@user-qy9tu8cy2p]

Hi, Thanks for your very knowledgeable videos. Please suggest which DL model/video should I follow for crop detection and identification in field?

[@atultiwari88]

wow.. thank you very much sir. you read my mind. I have been searching here and there for quite some time now abut how I can use YOLO or any deep learning trained model into actual Whole Slide Imaging, but I am not able to find any satisfactory answer so far. I hope in your next video I would get the answer I am looking for. This 2 weeks wait is killing me. please release the next video early.

[@DigitalSreeni]

:)

[@joshualee6559]

I'd be really interested in a video about 3D MRI segmentation using these techniques.

[@Koldim2001]

Please make a video on this topic - patch based inference

[@umairsabir3519]

Can you please train a panoptic segmentation using custom dataset. Also which one is better, detrctron 2 or google's paniptic algorithm ?

[@xd_ut-boi6698]

Would love to see some projects on genomics ✌🏻

[@DigitalSreeni]

Any specific type of project your are looking for? Can you please point me towards a specific dataset so I understand the challenge to be solved?

[@user-et2xr1kw5c]

Is there a chance to get processed files, that have gone through the all initial steps up until model itself?

[@vy1w1yv]

Which performed best on the dataset, YOLOv8 or detectron 2?

[@sayyamhanamannavar6413]

Hello sir I want to start learning about image processing and python and implement it on fashion runways to analyse fashion trends ..could you please tell me what all concepts do I need to cover in order to be able to do that.

[@anammanzoor1166]

Hi can you please make one video on initial setup in venv for coding which version you have used ?

[@chouchou2445]

hi please i need help i'ma one of your followers and i use apeer to annotate my dataset, but i can't export or even save on line annotations i create please help it's an emergency

[@palurikrishnaveni8344]

If my dataset does not have a separate mask-labeled dataset, how can I implement it? For the UNET case also, I am suffering from the problem. Can you show the code for any dataset with no mask dataset?

[@hoinhtrieu9637]

separate mask-labeled for instance segmentation, and binary mask for sematic segmentation. Based on your label type, the problem approach is totally different.

[@user-mb5tq8du1f]

can i apply the same method for oral cancer dataset

[@user-tv6eq4zc5e]

where does he use the validation dataset? He registered it, and didn't use it anywhere - why?

[@NISHATZAMAN-cv3qo]

how can I use Stl files for segmentation?? it doesn't have any mask. how can I put this in my model?

[@sudiptaproggaislam7321]

please make a video of detectron2 for panoptic segmentation...

[@cherryblxssxm.]

Hi, Thanks for your very useful videos. Please take some examples to use CNN-Transformer Hybrid model in Medical images analysis (ViT models) . Thanks a lot. Good Luck. M. Rezaei

[@amiralioghli8622]

Hi sir, thank you for sharing your valuable information with us. I have a suggestion. My comment does not pertain to the content of the current video since you've uploaded new content. Therefore, I am writing it here for you to read. I recently completed all your tutorials on time series data, both univariate and multivariate. From what I gathered in your tutorials, your primary focus lies in image and NLP. Hence, I'd like to request, if possible, Please make a series on implementing Transformers in time series data. This could cover classification, forecasting, or anomaly detection-just one of these aspects would be sufficient for educational purposes. I'm interested in understanding the methodology you propose in dealing with this. Currently, Transformers are extensively used in sequential data and image segmentation. Surprisingly, there's a scarcity of tutorials available on KZbin that delve into employing Transformers specifically for time series data.

[@DigitalSreeni]

My expertise lies in image processing which is why I focus primarily on that topic. Although I should mention that I experiment with financial data quite a bit and probably should find time to record some tutorials. To answer your question - I am not convinced about Transformers being the right tools for time series. I consistently see better results using the traditional AR models.

[@samarafroz9852]

Please made tutorial on zero Shot medical image segmentation Sir

[@vigneshvicky6720]

We need theory, architecture , loss ext... Code we can find from net itself

[@DigitalSreeni]

This video is a specific project focused. I have covered the basics of Detectron2 and YOLO in separate videos. For example, here is the Detectron2 intro video link: kzbin.info/www/bejne/gHqzk5yifd15mKs

[@MrEri299]

where is code?

[@user-qp7jz2ih9j]

Please make video on Image Segmentation using Zernike Moments #Zernike_Moments

[@zahraalidousti]

hi, I want to ask you about an error related to openslide library. Would you please give me your email to send it? Thank you

[@Koldim2001]

Please make a video on this topic - patch-based inference