Annotate Images Like a Pro: Python Image Annotation Tool Demo

Image Annotation Made Easy with DigitalSreeni's Python Tool:

In this video, I walk you through my Python-based image annotation application and its associated tools, providing a step-by-step demo to help you get started.

Topics Covered:

- Creating new projects and adding 2D and multi-dimensional images (TIFF, CZI).
- Searching for projects based on metadata
- Manual annotation of 2D images and slices from multi-dimensional images using polygon and rectangle tools.
- Semi-automatic annotations with the Segment Anything Model (SAM).
- Renaming and assigning colors to classes for better organization.
- Exporting annotations to various formats: COCO JSON, YOLO v8, labeled images, semantic images, Pascal VOC bounding boxes.
- Training custom model using YOLO
- Using the trained model to predict on images and converting predictions to annotations

Additional Tools:
- Annotation statistics
- Combining JSON annotations
- Data splitting
- Patch extraction
- Data augmentation of images and annotations

Links:

To Install:
pip install digitalsreeni-image-annotator

Once installed, simply type sreeni in your command prompt within the correct environment to launch the application.

You can download SAM models from the following links. Please be cautious about the large model on systems with limited memory.

It is recommended to place the SAM models in a directory from where you normally start the application to avoid multiple downloads of the same models from the Ultralytics server.

[texasfossilguy]

Wow this is incredible! THANK YOU SO MUCH!!!! I seriously cannot express how helpful this software is. This is exactly what I have been needing for years!

[datapro]

Before you released this I was using SAM, labelimg etc etc. Your tool is just great - it brings everything together in one place. Many thanks Sreeni.

[RithikPurohit]

Found this exactly when I needed it.
This is amazing and so simple. Thank you very much.

[mmrsagar]

Excellent tool. I am really surprised how well this works. I know there are lots of people who would be benefited by your tool without knowing much coding.
I think it would also be very beneficial for a lot of people like me who also wants to build some gui applications like yours. If you can make the step by step tutorials for how to package the application and push it to pypl and a bit about making the gui application.

[tanvikumari]

Great Work Sir. Sam model blow the mind. Thank you! Sir.

[qwiksponge]

Hi Sreeni, I absolutely love your annotation tool. I found it a bit hard to navigate the photos using the sliding bars (especially after I zoomed in). Could you kindly add a function that allows mouse right click to drag around the photo? Much appreciated!

[mohitgupta]

Hey, thanks for this video and for sharing the tool! It looks very useful to me. I will definitely try it and share the experience!

[ronaldoleoni]

First of all, I want to thank you for providing such a helpful tool. It's really helping me with my annotations, but I want to say something that should be useful in the next versions: I'm facing trouble right now with Detectron2 and the COCO exported annotations because the bounding boxes exported by the image annotator are (XYHW) instead of (XYXY). Maybe picking one of these in the export could be very useful. Again, thank you, it's an incredible tool, SAM helps a lot.

[TanTran-ferf]

I'm a newbie with Detectron2 and Annotation. How can I annotate a list of over 1000 images in Coco format without manually bounding box and labeling manually?

[chaospaypal]

There's a tool out there (SAM only) which does not use a drawn rectangle, but instead you click in the middle of an object and it uses SAM to find the boundaries. Do you think something like that makes sense here, is that possible? Basically what you do is if you find something interesting (like a mitrochondria) you just click on it.

[kannalokesh]

Great tool, if it has integration of cloud services for data storage, then it will be a nice tool

[peterslater]

Excellent tool. Thank you. Only one thing that for me is that I want try unet with masks. Is there a quick way to read and use annotations to form a binary file.

[VikashKumar-tyuy]

Great Work, that's what i was looking for.
One question how can we use this application as an api so that a larger user can use this app for my specific dataset

[kashif]

I get 8 values for a bounding box in annotation file file even when I select rectangle.

[HusnaUmeran]

Thank you for the video it's super helpful. Can you also talk about how you trained yolo11?

[strongcourage-ce]

It seems not use my GPU. And the SAM always can't select correct area but some small points or region.

[brunospfc]

Great tool, i'm using to visualize datasets, sure it needs some tunning...i would suggest option to "fix the default zoom", as i have to zoom out of every single image(4k), and zoom in the region the pointer is....and some tutorial on importing datasets (Yolo, what structure the program expects in the actual version...)...anyway, thanks for the free work, it is certanly going in the right direction !

[MansiBansal-juvc]

I want the annotated images to be saved as such but instead they get saved as coordinates that is I want to open them with labelling, even outside this annotator, pls help me out Sir. I have a presentation to give.

[CarloSantos-vs]

Can I use this as alternative to roboflow?

[Laura-cbe]

Hi, this annotation tool looks brilliant! I wondered if it is possible to use the paintbrush tool to create contour segmentations rather than area-based ones? e.g. for a doughnut shape, can we segment just the circular contour rather than an area-filled circle?