ConvNeXt V2 tiny design
According to the model card, this tiny ConvNeXt V2 model uses the fully convolutional masked autoencoder framework and a Global Response Normalization layer.
Open Source Model Profile · facebook
convnextv2-tiny-22k-384
convnextv2-tiny-22k-384 is a 28.64M-parameter ConvNeXt V2 image classifier from facebook. According to the model card, it was fine-tuned on ImageNet-22K at 384x384.
- Publisher
- Task
- image-classification
- Model type
- convnextv2
- License
- apache-2.0
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
convnextv2-tiny-22k-384 is published by facebook as an image-classification model. The captured configuration identifies ConvNextV2ForImageClassification with model type convnextv2, and Safetensors metadata reports 28,635,496 parameters. According to the model card, it was pretrained with FCMAE and fine-tuned on ImageNet-22K at 384x384.
Recorded capabilities
ImageNet-22K at 384x384
According to the model card, the model was pretrained with FCMAE and fine-tuned on ImageNet-22K at 384x384, consistent with its dataset:imagenet-22k tag.
Transformers-compatible release
Hub metadata identifies transformers library support with pytorch, tf, safetensors, convnextv2, vision, and endpoints compatibility.
Documented classification workflow
According to the model card, the raw model is for image classification, with an example classifying COCO or cats-image inputs into 1,000 ImageNet classes.
Use cases in the source record
- Image classification of inputs into one of the 1,000 ImageNet classes using the model card's documented Transformers workflow.
- Convolutional-architecture study referencing the card's FCMAE pretraining and ImageNet-22K fine-tuning description.
Limitations and unknowns
- No independently measured evaluation results were extracted from this record.
- No training hyperparameters, compute, or dataset size beyond ImageNet-22K naming were extracted.
- According to the card disclaimer, the releasing team did not write the model card; the Hugging Face team wrote it instead.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: facebook/convnextv2-tiny-22k-384
Captured: Unknown. Processed: 2026-09-07T19:34:44.310396+00:00.
ConvNeXt V2 (tiny-sized model) ConvNeXt V2 model pretrained using the FCMAE framework and fine-tuned on the ImageNet-22K dataset at resolution 384x384. It was introduced in the paper ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders by Woo et al. and first released in this repository . Disclaimer: The team releasing ConvNeXT V2 did not write a model card for this model so this model card has been written by the Hugging Face team. Model description ConvNeXt V2 is a pure convolutional model (ConvNet) that introduces a fully convolutional masked autoencoder framework (FCMAE) and a new Global Response Normalization…
F001F002F003F004F005F006F007F008F009F010F011F012F013F014F015F016