You can deploy this model and start earning money today!

kerasformers/
mobilenetv3_large_100_ra4_e3600_r224_in1k

Name: mobilenetv3_large_100_ra4_e3600_r224_in1k
Availability: InStock

timm/mobilenetv3_large_100.ra4_e3600_r224_in1k

finetune

apache-2.0

Views

No views yet

See our collection for all versions of MobileNetV3.

Run MobileNetV3 with Keras 3: JAX, PyTorch, or TensorFlow

GitHub

zeromodels/mobilenetv3_large_100_ra4_e3600_r224_in1k

Paper: Searching for MobileNetV3 (arXiv:1905.02244) · HF Papers

MobileNetV3 combines NAS, NetAdapt, and hard-swish / SE for efficient mobile vision. Classifier or backbone.

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

Pure-Keras 3 conversion of timm/mobilenetv3_large_100.ra4_e3600_r224_in1k for zeromodels. One implementation runs unmodified on TensorFlow / Torch / JAX.

This is an image-classification / backbone checkpoint (MobileNetV3ImageClassify / MobileNetV3Model).

✨ Quick start

python

import os

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

from PIL import Image
from zeromodels.models.mobilenetv3 import MobileNetV3ImageClassify, MobileNetV3Model, MobileNetV3ImageProcessor

model = MobileNetV3ImageClassify.from_weights("zeromodels/mobilenetv3_large_100_ra4_e3600_r224_in1k")
processor = MobileNetV3ImageProcessor.from_weights("zeromodels/mobilenetv3_large_100_ra4_e3600_r224_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 = MobileNetV3Model.from_weights("zeromodels/mobilenetv3_large_100_ra4_e3600_r224_in1k", as_backbone=True)
features = backbone(pixels, training=False)

Load any MobileNetV3 variant the same way with from_weights("zeromodels/<variant>"):

Variant	Hub
`mobilenetv3_large_100_miil_in21k`	`zeromodels/mobilenetv3_large_100_miil_in21k`
`mobilenetv3_large_100_miil_in21k_ft_in1k`	`zeromodels/mobilenetv3_large_100_miil_in21k_ft_in1k`
`mobilenetv3_large_100_ra4_e3600_r224_in1k`	`zeromodels/mobilenetv3_large_100_ra4_e3600_r224_in1k`
`mobilenetv3_large_100_ra_in1k`	`zeromodels/mobilenetv3_large_100_ra_in1k`
`mobilenetv3_large_150d_ra4_e3600_r256_in1k`	`zeromodels/mobilenetv3_large_150d_ra4_e3600_r256_in1k`
`mobilenetv3_rw_rmsp_in1k`	`zeromodels/mobilenetv3_rw_rmsp_in1k`
`mobilenetv3_small_050_lamb_in1k`	`zeromodels/mobilenetv3_small_050_lamb_in1k`
`mobilenetv3_small_075_lamb_in1k`	`zeromodels/mobilenetv3_small_075_lamb_in1k`
`mobilenetv3_small_100_lamb_in1k`	`zeromodels/mobilenetv3_small_100_lamb_in1k`

Tips

Set KERAS_BACKEND before importing Keras / zeromodels.
MobileNetV3ImageClassify returns class logits; MobileNetV3Model returns features (as_backbone=True for multi-scale stages).
See docs and Loading Weights.
Upstream / timm checkpoints: MobileNetV3ImageClassify.from_weights("hf:timm/mobilenetv3_large_100.ra4_e3600_r224_in1k").

Special Thanks

A huge thank you to the MobileNetV3 authors and the timm / Hub communities for creating and releasing these models.

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

kerasformers/mobilenetv3_large_100_ra4_e3600_r224_in1k