zeromodels/tf_efficientnetv2_l_in1k

See our collection for all versions of EfficientNetV2.

Run EfficientNetV2 with Keras 3: JAX, PyTorch, or TensorFlow

zeromodels/tf_efficientnetv2_l_in1k

Paper: EfficientNetV2: Smaller Models and Faster Training (arXiv:2104.00298) · HF Papers

EfficientNetV2 trains faster with Fused-MBConv and progressive learning. Same ImageClassify / Model API as EfficientNet.

For more details on the model, please go to the upstream model card.

Pure-Keras 3 conversion of timm/tf_efficientnetv2_l.in1k for zeromodels. One implementation runs unmodified on TensorFlow / Torch / JAX.

This is an image-classification / backbone checkpoint (EfficientNetV2ImageClassify / EfficientNetV2Model).

✨ Quick start

import os

os.environ["KERAS_BACKEND"] = "torch"  # or "jax" / "tensorflow"

from PIL import Image
from zeromodels.models.efficientnetv2 import EfficientNetV2ImageClassify, EfficientNetV2Model, EfficientNetV2ImageProcessor

model = EfficientNetV2ImageClassify.from_weights("zeromodels/tf_efficientnetv2_l_in1k")
processor = EfficientNetV2ImageProcessor.from_weights("zeromodels/tf_efficientnetv2_l_in1k")

image = Image.open("your_image.jpg").convert("RGB")
pixels = processor(image)  # resize + normalize (normalization lives in the processor)
logits = model(pixels, training=False)
print(logits.shape)  # (1, num_classes)

# Feature extraction: the backbone without the classifier head
backbone = EfficientNetV2Model.from_weights("zeromodels/tf_efficientnetv2_l_in1k", as_backbone=True)
features = backbone(pixels, training=False)

Load any EfficientNetV2 variant the same way with from_weights("zeromodels/<variant>"):

Variant	Hub
tf_efficientnetv2_b0_in1k	zeromodels/tf_efficientnetv2_b0_in1k
tf_efficientnetv2_b1_in1k	zeromodels/tf_efficientnetv2_b1_in1k
tf_efficientnetv2_b2_in1k	zeromodels/tf_efficientnetv2_b2_in1k
tf_efficientnetv2_b3_in1k	zeromodels/tf_efficientnetv2_b3_in1k
tf_efficientnetv2_b3_in21k_ft_in1k	zeromodels/tf_efficientnetv2_b3_in21k_ft_in1k
tf_efficientnetv2_l_in1k	zeromodels/tf_efficientnetv2_l_in1k
tf_efficientnetv2_l_in21k	zeromodels/tf_efficientnetv2_l_in21k
tf_efficientnetv2_l_in21k_ft_in1k	zeromodels/tf_efficientnetv2_l_in21k_ft_in1k
tf_efficientnetv2_m_in1k	zeromodels/tf_efficientnetv2_m_in1k
tf_efficientnetv2_m_in21k	zeromodels/tf_efficientnetv2_m_in21k
tf_efficientnetv2_m_in21k_ft_in1k	zeromodels/tf_efficientnetv2_m_in21k_ft_in1k
tf_efficientnetv2_s_in1k	zeromodels/tf_efficientnetv2_s_in1k
tf_efficientnetv2_s_in21k	zeromodels/tf_efficientnetv2_s_in21k
tf_efficientnetv2_s_in21k_ft_in1k	zeromodels/tf_efficientnetv2_s_in21k_ft_in1k
tf_efficientnetv2_xl_in21k	zeromodels/tf_efficientnetv2_xl_in21k
tf_efficientnetv2_xl_in21k_ft_in1k	zeromodels/tf_efficientnetv2_xl_in21k_ft_in1k

Tips

• Set KERAS_BACKEND before importing Keras / zeromodels.
• EfficientNetV2ImageClassify returns class logits; EfficientNetV2Model returns features (as_backbone=True for multi-scale stages).
• See docs and Loading Weights.
• Upstream / timm checkpoints: EfficientNetV2ImageClassify.from_weights("hf:timm/tf_efficientnetv2_l.in1k").

Special Thanks

A huge thank you to the EfficientNetV2 authors and the timm / Hub communities for creating and releasing these models.

License: see YAML license (usually matches the upstream checkpoint).