Integration of YOLO-NAS into FiftyOne Model Zoo #3925

[AdonaiVera]

Hi FiftyOne Team 👋

We're excited to announce the successful integration of YOLO-NAS into the FiftyOne Model Zoo, a collaborative effort with @jacobmarks. This addition enriches our model offerings with advanced vision capabilities from YOLO-NAS.

Integration Highlights:

YOLO-NAS, sourced from SuperGradients' GitHub, is now seamlessly integrated into FiftyOne, offering users its state-of-the-art vision capabilities within the familiar FiftyOne environment.
This integration is designed for smooth compatibility and user-friendly interaction with existing FiftyOne features.

Interactive Demonstration:

I've developed an informative Jupyter notebook to illustrate YOLO-NAS's functions in FiftyOne, focusing on its practical applications in vision tasks. Check out the FiftyOne + YOLO-NAS Demo for a hands-on experience.

Besides, you can import the model directly from super_gradients library or Fiftyone zoo models

Example Usage

import fiftyone as fo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart", max_samples=10)

Detection with super_gradients

from super_gradients.training import models

model = models.get("yolo_nas_m", pretrained_weights="coco")
dataset.apply_model(model, label_field="yolo-nas", confidence_thresh=0.7)

Detection with FiftyOne zoo model

model = foz.load_zoo_model("yolo-nas-torch)
dataset.apply_model(model, label_field="yolo-nas", confidence_thresh=0.7)

YOLO-NAS Model Variants

The integration includes a variety of YOLO-NAS models, each with its specific application potential, detailed in the demo notebook.

Your insights and suggestions are invaluable. Please feel free to share your thoughts or seek clarification regarding this integration.

Warm regards,
Adonai Vera

[brimoor]

@AdonaiVera this is awesome!

Looks like there's a merge conflict with the latest develop branch; can you rebase/merge onto the latest develop please?

[AdonaiVera]

Hi @brimoor
Sure, solved it 👍

[brimoor]

LGTM from a code standpoint! 🔥

@jacobmarks I haven't tried to run the model myself, can you confirm that it works for you before we merge?

[codecov]

Codecov Report

Attention: 39 lines in your changes are missing coverage. Please review.

Comparison is base (f306890) 15.84% compared to head (20704f3) 15.84%.
Report is 14 commits behind head on develop.

Files	Patch %	Lines
.../src/components/Sidebar/Entries/PathValueEntry.tsx	0.00%	36 Missing ⚠️
app/packages/state/src/recoil/schema.ts	25.00%	3 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #3990   +/-   ##
========================================
  Coverage    15.84%   15.84%           
========================================
  Files          731      731           
  Lines        81907    81907           
  Branches      1093     1093           
========================================
  Hits         12980    12980           
  Misses       68927    68927           

Flag	Coverage Δ
app	15.84% <2.50%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jacobmarks]

Amazing work @AdonaiVera !!! The kudos is ALL YOURS here. What an incredible addition to the library/ecosystem :)

@brimoor I've verified that both options for applying the model work 👍 💪

[jacobmarks]

@AdonaiVera we should also update the docstrings for apply_model() to add the new option :)

[brimoor]

I went ahead and added the documentation to apply_model() ✅

[AdonaiVera]

Amazing, thank you @brimoor and @jacobmarks 💪

The next step is going to be Yolo-nas-pose 🚀