# MobileNet-V3

**Category:** Efficient/Mobile · **Code:**
[`timm/models/mobilenetv3.py`](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/mobilenetv3.py)
· **Architectures:** 33 · **Pretrained weights:** 45 (39 with ImageNet-1k results)

MobileNet-V3 and the related networks built on the same `MobileNetV3` class: MobileNet-V4, LCNet (PP-LCNet), FBNet-V3
and TensorFlow ported MobileNet-V3 weights (`tf_*`).

## Papers

* [Searching for MobileNetV3](https://arxiv.org/abs/1905.02244)
* [FBNetV3: Joint Architecture-Recipe Search using Predictor Pretraining](https://arxiv.org/abs/2006.02049)
* [PP-LCNet: A Lightweight CPU Convolutional Neural Network](https://arxiv.org/abs/2109.15099)
* [MobileNetV4 -- Universal Models for the Mobile Ecosystem](https://arxiv.org/abs/2404.10518)
* [ImageNet-21K Pretraining for the Masses](https://arxiv.org/abs/2104.10972)

## Usage

See the [Quickstart](../quickstart) for loading pretrained weights, image preprocessing and fine-tuning, and [Feature
Extraction](../feature_extraction) for using the models as backbones.

```py
>>> import timm
>>> model = timm.create_model('mobilenetv4_conv_aa_large.e230_r384_in12k_ft_in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `MobileNetV3` |
|---|---|
| [`features_only=True` feature maps](../model_api#intermediate-features) | ✓ (strides 2, 4, 8, 16, 32) |
| [`forward_intermediates()`](../model_api#intermediate-features) | ✓ |
| [`prune_intermediate_layers()`](../model_api#intermediate-features) | ✓ |
| [Any input size, no changes needed](../model_api#input-size) | ✓ |
| [`set_input_size()`](../model_api#input-size) | – |
| [`dynamic_img_size=True`](../model_api#input-size) | – |
| [`set_grad_checkpointing()`](../model_api#training-hooks) | ✓ |
| [Layer-wise LR decay (`group_matcher()`)](../model_api#training-hooks) | ✓ |
| [`reset_classifier()`](../model_api#classifier) | ✓ |

✓ supported, ✓ (square) works for square sizes only, ✗ not supported, – not needed (the model already accepts any input
size). Checked by running each feature on a randomly initialized model (`MobileNetV3` with `mobilenetv3_small_050` at
224), other architectures of the same class are expected to behave the same. See the [Model API](../model_api) page for
details.

## Architectures

| Architecture | Params (M) | GMACs | Input size | Pretrained weights |
|---|---:|---:|---:|---:|
| `fbnetv3_b` | 8.60 | 0.42 | 224 | 1 |
| `fbnetv3_d` | 10.31 | 0.52 | 224 | 1 |
| `fbnetv3_g` | 16.62 | 1.28 | 240 | 1 |
| `lcnet_035` | 1.64 | 0.03 | 224 | 0 |
| `lcnet_050` | 1.88 | 0.05 | 224 | 1 |
| `lcnet_075` | 2.36 | 0.10 | 224 | 1 |
| `lcnet_100` | 2.95 | 0.16 | 224 | 1 |
| `lcnet_150` | 4.50 | 0.34 | 224 | 0 |
| `mobilenetv3_large_075` | 3.99 | 0.16 | 224 | 0 |
| `mobilenetv3_large_100` | 5.48 | 0.23 | 224 | 4 |
| `mobilenetv3_large_150d` | 14.62 | 1.03 | 256 | 1 |
| `mobilenetv3_rw` | 5.48 | 0.23 | 224 | 1 |
| `mobilenetv3_small_050` | 1.59 | 0.03 | 224 | 1 |
| `mobilenetv3_small_075` | 2.04 | 0.05 | 224 | 1 |
| `mobilenetv3_small_100` | 2.54 | 0.06 | 224 | 1 |
| `mobilenetv4_conv_aa_large` | 32.59 | 9.63 | 448 | 4 |
| `mobilenetv4_conv_aa_medium` | 9.72 | 1.58 | 256 | 0 |
| `mobilenetv4_conv_blur_medium` | 9.72 | 1.22 | 224 | 1 |
| `mobilenetv4_conv_large` | 32.59 | 6.43 | 384 | 2 |
| `mobilenetv4_conv_medium` | 9.72 | 1.10 | 256 | 6 |
| `mobilenetv4_conv_small` | 3.77 | 0.19 | 224 | 3 |
| `mobilenetv4_conv_small_035` | 1.91 | 0.05 | 224 | 0 |
| `mobilenetv4_conv_small_050` | 2.24 | 0.07 | 224 | 1 |
| `mobilenetv4_hybrid_large` | 37.76 | 7.77 | 384 | 2 |
| `mobilenetv4_hybrid_large_075` | 22.75 | 2.06 | 256 | 0 |
| `mobilenetv4_hybrid_medium` | 11.07 | 1.29 | 256 | 5 |
| `mobilenetv4_hybrid_medium_075` | 7.31 | 0.66 | 224 | 0 |
| `tf_mobilenetv3_large_075` | 3.99 | 0.16 | 224 | 1 |
| `tf_mobilenetv3_large_100` | 5.48 | 0.23 | 224 | 1 |
| `tf_mobilenetv3_large_minimal_100` | 3.92 | 0.22 | 224 | 1 |
| `tf_mobilenetv3_small_075` | 2.04 | 0.05 | 224 | 1 |
| `tf_mobilenetv3_small_100` | 2.54 | 0.06 | 224 | 1 |
| `tf_mobilenetv3_small_minimal_100` | 2.04 | 0.06 | 224 | 1 |

GMACs and parameter counts are for the default input size and classifier of each architecture, from the [benchmark
results](https://github.com/huggingface/pytorch-image-models/tree/main/results).

## Pretrained weights

Top-1 is ImageNet-1k validation accuracy from
[`results-imagenet.csv`](https://github.com/huggingface/pytorch-image-models/blob/main/results/results-imagenet.csv) at
the listed eval size, - if the weight has no ImageNet-1k result (e.g. pretraining or image-text weights). Classes is the
size of the pretrained classifier, none for weights without one.

Weights are grouped by source: the paper or origin given in the pretrained config, else a Hub organization other than
`timm`, else the first part of the pretrained tag (`architecture.tag`). Sources with a single weight are listed under
other.

### `ra2` tags

| Weight | Input size | Top-1 (%) | Eval size | Classes | License | Hub |
|---|---:|---:|---:|---:|---|---|
| `fbnetv3_b.ra2_in1k` | 224 (test 256) | 79.154 | 256 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/fbnetv3_b.ra2_in1k) |
| `fbnetv3_d.ra2_in1k` | 224 (test 256) | 79.682 | 256 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/fbnetv3_d.ra2_in1k) |
| `fbnetv3_g.ra2_in1k` | 240 (test 288) | 82.034 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/fbnetv3_g.ra2_in1k) |
| `lcnet_050.ra2_in1k` | 224 | 63.138 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/lcnet_050.ra2_in1k) |
| `lcnet_075.ra2_in1k` | 224 | 68.786 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/lcnet_075.ra2_in1k) |
| `lcnet_100.ra2_in1k` | 224 | 72.130 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/lcnet_100.ra2_in1k) |

### `ix` tags

| Weight | Input size | Top-1 (%) | Eval size | Classes | License | Hub |
|---|---:|---:|---:|---:|---|---|
| `mobilenetv4_hybrid_large.ix_e600_r384_in1k` | 384 (test 448) | 84.352 | 448 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_hybrid_large.ix_e600_r384_in1k) |
| `mobilenetv4_hybrid_medium.ix_e550_r256_in1k` | 256 (test 320) | 82.500 | 320 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_hybrid_medium.ix_e550_r256_in1k) |
| `mobilenetv4_hybrid_medium.ix_e550_r384_in1k` | 384 (test 448) | 83.390 | 448 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_hybrid_medium.ix_e550_r384_in1k) |

### `lamb` tags

| Weight | Input size | Top-1 (%) | Eval size | Classes | License | Hub |
|---|---:|---:|---:|---:|---|---|
| `mobilenetv3_small_050.lamb_in1k` | 224 | 57.918 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv3_small_050.lamb_in1k) |
| `mobilenetv3_small_075.lamb_in1k` | 224 | 65.284 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv3_small_075.lamb_in1k) |
| `mobilenetv3_small_100.lamb_in1k` | 224 | 67.636 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv3_small_100.lamb_in1k) |

### `ra4` tags

| Weight | Input size | Top-1 (%) | Eval size | Classes | License | Hub |
|---|---:|---:|---:|---:|---|---|
| `mobilenetv3_large_100.ra4_e3600_r224_in1k` | 224 (test 256) | 77.182 | 256 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv3_large_100.ra4_e3600_r224_in1k) |
| `mobilenetv3_large_150d.ra4_e3600_r256_in1k` | 256 (test 320) | 81.832 | 320 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv3_large_150d.ra4_e3600_r256_in1k) |

### ImageNet-21K Pretraining for the Masses

Pretrained configs reference [ImageNet-21K Pretraining for the Masses](https://arxiv.org/abs/2104.10972).

| Weight | Input size | Top-1 (%) | Eval size | Classes | License | Hub |
|---|---:|---:|---:|---:|---|---|
| `mobilenetv3_large_100.miil_in21k` | 224 | - | - | 11221 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv3_large_100.miil_in21k) |
| `mobilenetv3_large_100.miil_in21k_ft_in1k` | 224 | 77.920 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv3_large_100.miil_in21k_ft_in1k) |

### Other

| Weight | Input size | Top-1 (%) | Eval size | Classes | License | Hub |
|---|---:|---:|---:|---:|---|---|
| `mobilenetv3_large_100.ra_in1k` | 224 | 75.784 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv3_large_100.ra_in1k) |
| `mobilenetv3_rw.rmsp_in1k` | 224 | 75.616 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv3_rw.rmsp_in1k) |
| `mobilenetv4_conv_aa_large.e230_r384_in12k` | 384 (test 448) | - | - | 11821 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_aa_large.e230_r384_in12k) |
| `mobilenetv4_conv_aa_large.e230_r384_in12k_ft_in1k` | 384 (test 480) | 84.772 | 480 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_aa_large.e230_r384_in12k_ft_in1k) |
| `mobilenetv4_conv_aa_large.e230_r448_in12k_ft_in1k` | 448 (test 544) | 84.990 | 544 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_aa_large.e230_r448_in12k_ft_in1k) |
| `mobilenetv4_conv_aa_large.e600_r384_in1k` | 384 (test 480) | 83.826 | 480 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_aa_large.e600_r384_in1k) |
| `mobilenetv4_conv_blur_medium.e500_r224_in1k` | 224 (test 256) | 80.152 | 256 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_blur_medium.e500_r224_in1k) |
| `mobilenetv4_conv_large.e500_r256_in1k` | 256 (test 320) | 82.672 | 320 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_large.e500_r256_in1k) |
| `mobilenetv4_conv_large.e600_r384_in1k` | 384 (test 448) | 83.418 | 448 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_large.e600_r384_in1k) |
| `mobilenetv4_conv_medium.e180_ad_r384_in12k` | 384 | - | - | 11821 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_medium.e180_ad_r384_in12k) |
| `mobilenetv4_conv_medium.e180_r384_in12k` | 384 | - | - | 11821 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_medium.e180_r384_in12k) |
| `mobilenetv4_conv_medium.e250_r384_in12k` | 384 | - | - | 11821 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_medium.e250_r384_in12k) |
| `mobilenetv4_conv_medium.e250_r384_in12k_ft_in1k` | 384 | 82.776 | 384 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_medium.e250_r384_in12k_ft_in1k) |
| `mobilenetv4_conv_medium.e500_r224_in1k` | 224 (test 256) | 79.836 | 256 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_medium.e500_r224_in1k) |
| `mobilenetv4_conv_medium.e500_r256_in1k` | 256 (test 320) | 80.890 | 320 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_medium.e500_r256_in1k) |
| `mobilenetv4_conv_small.e1200_r224_in1k` | 224 (test 256) | 74.282 | 256 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_small.e1200_r224_in1k) |
| `mobilenetv4_conv_small.e2400_r224_in1k` | 224 (test 256) | 74.644 | 256 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_small.e2400_r224_in1k) |
| `mobilenetv4_conv_small.e3600_r256_in1k` | 256 (test 320) | 75.616 | 320 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_small.e3600_r256_in1k) |
| `mobilenetv4_conv_small_050.e3000_r224_in1k` | 224 (test 256) | 65.826 | 256 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_conv_small_050.e3000_r224_in1k) |
| `mobilenetv4_hybrid_large.e600_r384_in1k` | 384 (test 448) | 84.264 | 448 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_hybrid_large.e600_r384_in1k) |
| `mobilenetv4_hybrid_medium.e200_r256_in12k` | 256 (test 320) | - | - | 11821 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_hybrid_medium.e200_r256_in12k) |
| `mobilenetv4_hybrid_medium.e200_r256_in12k_ft_in1k` | 256 (test 320) | 82.998 | 320 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_hybrid_medium.e200_r256_in12k_ft_in1k) |
| `mobilenetv4_hybrid_medium.e500_r224_in1k` | 224 (test 256) | 81.284 | 256 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mobilenetv4_hybrid_medium.e500_r224_in1k) |
| `tf_mobilenetv3_large_075.in1k` | 224 | 73.442 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/tf_mobilenetv3_large_075.in1k) |
| `tf_mobilenetv3_large_100.in1k` | 224 | 75.510 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/tf_mobilenetv3_large_100.in1k) |
| `tf_mobilenetv3_large_minimal_100.in1k` | 224 | 72.270 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/tf_mobilenetv3_large_minimal_100.in1k) |
| `tf_mobilenetv3_small_075.in1k` | 224 | 65.706 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/tf_mobilenetv3_small_075.in1k) |
| `tf_mobilenetv3_small_100.in1k` | 224 | 67.922 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/tf_mobilenetv3_small_100.in1k) |
| `tf_mobilenetv3_small_minimal_100.in1k` | 224 | 62.892 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/tf_mobilenetv3_small_minimal_100.in1k) |