# Res2Net

**Category:** CNN · **Code:**
[`timm/models/res2net.py`](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/res2net.py) ·
**Architectures:** 9 · **Pretrained weights:** 9 (9 with ImageNet-1k results)

Res2Net splits the channels of the bottleneck block into groups connected hierarchically, giving several receptive field
sizes within one block. Includes Res2NeXt.

## Papers

* [Res2Net: A New Multi-scale Backbone Architecture](https://arxiv.org/abs/1904.01169)

## 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('res2net50_14w_8s.in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `ResNet` |
|---|---|
| [`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 (`ResNet` with `res2next50` 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 |
|---|---:|---:|---:|---:|
| `res2net50_14w_8s` | 25.06 | 4.21 | 224 | 1 |
| `res2net50_26w_4s` | 25.70 | 4.28 | 224 | 1 |
| `res2net50_26w_6s` | 37.05 | 6.33 | 224 | 1 |
| `res2net50_26w_8s` | 48.40 | 8.37 | 224 | 1 |
| `res2net50_48w_2s` | 25.29 | 4.18 | 224 | 1 |
| `res2net50d` | 25.72 | 4.52 | 224 | 1 |
| `res2net101_26w_4s` | 45.21 | 8.10 | 224 | 1 |
| `res2net101d` | 45.23 | 8.35 | 224 | 1 |
| `res2next50` | 24.67 | 4.20 | 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.

| Weight | Input size | Top-1 (%) | Eval size | Classes | License | Hub |
|---|---:|---:|---:|---:|---|---|
| `res2net50_14w_8s.in1k` | 224 | 78.142 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/res2net50_14w_8s.in1k) |
| `res2net50_26w_4s.in1k` | 224 | 77.994 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/res2net50_26w_4s.in1k) |
| `res2net50_26w_6s.in1k` | 224 | 78.586 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/res2net50_26w_6s.in1k) |
| `res2net50_26w_8s.in1k` | 224 | 78.950 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/res2net50_26w_8s.in1k) |
| `res2net50_48w_2s.in1k` | 224 | 77.528 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/res2net50_48w_2s.in1k) |
| `res2net50d.in1k` | 224 | 80.280 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/res2net50d.in1k) |
| `res2net101_26w_4s.in1k` | 224 | 79.192 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/res2net101_26w_4s.in1k) |
| `res2net101d.in1k` | 224 | 81.236 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/res2net101d.in1k) |
| `res2next50.in1k` | 224 | 78.238 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/res2next50.in1k) |