ConvNeXt V2

ConvNeXt V2 is an image classifier with global response normalization.

Tasks: classification
Install: pip install libreyolo
Support tier: Supported, since v1.6.0. Supporting trainables: kept green in CI, features land opportunistically.
Upstream: ConvNeXt V2 by Meta AI, MIT code; CC-BY-NC-4.0 weights. Paper, source
Licenses: Code MIT, weights MIT code; CC-BY-NC-4.0 weights. Commercial use

Install

bash

pip install "libreyolo"

Predict

Python

from libreyolo import LibreYOLO, SAMPLE_IMAGE model = LibreYOLO("LibreConvNeXtV2atto-cls.pt", device="cpu")result = model(SAMPLE_IMAGE)print(result.probs)

The eight sizes run classification at 224 pixels. Official pretrained weights are CC-BY-NC-4.0.

Train

Training rebuilds the classification head for an ImageFolder dataset. cls_pw accepts values from 0 to 1 and defaults to 0. class_weights=True selects the alternative sample-normalized weighting; it cannot combine with cls_pw>0. Weighting settings must match when resuming.

Dataset setup describes the required training data.

Python

from libreyolo import LibreYOLO model = LibreYOLO("LibreConvNeXtV2atto-cls.pt", device="cpu")# "smoke10" auto-downloads; for your own data, pass a folder with train/ and val/ class subfolders.model.train(data="smoke10", epochs=1, device="cpu", workers=0)

Validate

Python

from libreyolo import LibreYOLO model = LibreYOLO("LibreConvNeXtV2atto-cls.pt", device="cpu")metrics = model.val(data="smoke10", workers=0)print(metrics)

Use the dataset format for this task. Validation explains the dataset requirements and returned metrics.

Export

Task	ONNX	TorchScript	ExecuTorch	TensorRT	OpenVINO	Paddle	MNN	RKNN	ncnn	TFLite	CoreML	Core AI
Classification	Classification to ONNX: supported	Classification to TorchScript: supported	Classification to ExecuTorch: supported	Classification to TensorRT: supported	Classification to OpenVINO: supported	Classification to Paddle: not supported	Classification to MNN: not supported	Classification to RKNN: not supported	Classification to ncnn: supported	Classification to TFLite: not supported	Classification to CoreML: not supported	Classification to Core AI: not supported

Python

from libreyolo import LibreYOLO, SAMPLE_IMAGE model = LibreYOLO("LibreConvNeXtV2atto-cls.pt", device="cpu")model.export(format="onnx")

Export setup lists format dependencies and loading exported artifacts.

Checkpoints

File	Weights license
Classification
LibreConvNeXtV2atto-cls.pt	cc-by-nc-4.0
LibreConvNeXtV2femto-cls.pt	cc-by-nc-4.0
LibreConvNeXtV2pico-cls.pt	cc-by-nc-4.0
LibreConvNeXtV2n-cls.pt	cc-by-nc-4.0
LibreConvNeXtV2t-cls.pt	cc-by-nc-4.0
LibreConvNeXtV2b-cls.pt	cc-by-nc-4.0
LibreConvNeXtV2l-cls.pt	cc-by-nc-4.0
LibreConvNeXtV2h-cls.pt	cc-by-nc-4.0

Every file above exists in the LibreYOLO org today and downloads on first use.

Licensing

Check the license on the Hugging Face repository of the specific weights you download. Every checkpoint in the LibreYOLO org carries one, and they are not always the same across a family. That repository is the authoritative source; the summary below describes what applied when this page was last verified.

This is a description of the licenses involved, not legal advice. If the answer matters commercially, read the licenses yourself and take your own counsel.

Original work: ConvNeXt V2, Meta AI
Upstream license: MIT code; CC-BY-NC-4.0 weights
Upstream source: github.com/facebookresearch/ConvNeXt-V2
LibreYOLO code: MIT
Weights: MIT code; CC-BY-NC-4.0 weights, republished at huggingface.co/LibreYOLO
Interpretation: The source code is MIT. Official pretrained weights are CC-BY-NC-4.0 and exclude commercial use.

Citation

@article{Woo2023ConvNeXtV2,
  title={ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders},
  author={Sanghyun Woo, Shoubhik Debnath, Ronghang Hu, Xinlei Chen, Zhuang Liu, In So Kweon and Saining Xie},
  year={2023},
  journal={arXiv preprint arXiv:2301.00808},
}

Copied from the authors' citation block at raw.githubusercontent.com/facebookresearch/ConvNeXt-V2/main/README.md.