tf_efficientnet_lite4_in1k

See our collection for all versions of EfficientNet-Lite.

Run EfficientNet-Lite with Keras 3: JAX, PyTorch, or TensorFlow

zeromodels/tf_efficientnet_lite4_in1k

Paper: EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks (arXiv:1905.11946) · HF Papers

EfficientNet-Lite is the mobile/EdgeTPU-friendly EfficientNet family (no squeeze-excite, ReLU6). Classifier or backbone.

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

Pure-Keras 3 conversion of timm/tf_efficientnet_lite4.in1k for zeromodels. One implementation runs unmodified on TensorFlow / Torch / JAX.

This is an image-classification / backbone checkpoint (EfficientNetLiteImageClassify / EfficientNetLiteModel).

✨ Quick start

import os

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

from PIL import Image
from zeromodels.models.efficientnet_lite import EfficientNetLiteImageClassify, EfficientNetLiteModel, EfficientNetLiteImageProcessor

model = EfficientNetLiteImageClassify.from_weights("zeromodels/tf_efficientnet_lite4_in1k")
processor = EfficientNetLiteImageProcessor.from_weights("zeromodels/tf_efficientnet_lite4_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 = EfficientNetLiteModel.from_weights("zeromodels/tf_efficientnet_lite4_in1k", as_backbone=True)
features = backbone(pixels, training=False)

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

Tips

• Set KERAS_BACKEND before importing Keras / zeromodels.
• EfficientNetLiteImageClassify returns class logits; EfficientNetLiteModel returns features (as_backbone=True for multi-scale stages).
• See docs and Loading Weights.
• Upstream / timm checkpoints: EfficientNetLiteImageClassify.from_weights("hf:timm/tf_efficientnet_lite4.in1k").

Special Thanks

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

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