# PVTv2

**Category:** ViT · **Code:**
[`timm/models/pvt_v2.py`](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/pvt_v2.py) ·
**Architectures:** 7 · **Pretrained weights:** 7 (7 with ImageNet-1k results)

Pyramid Vision Transformer v2, a hierarchical transformer with spatial-reduction attention (linear in the `_li`
variant), overlapping patch embeddings and convolutional feed-forward layers.

## Papers

* [PVT v2: Improved Baselines with Pyramid Vision Transformer](https://arxiv.org/abs/2106.13797)

## 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('pvt_v2_b2.in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `PyramidVisionTransformerV2` |
|---|---|
| [`features_only=True` feature maps](../model_api#intermediate-features) | ✓ (strides 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 (`PyramidVisionTransformerV2` with `pvt_v2_b0` 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 |
|---|---:|---:|---:|---:|
| `pvt_v2_b0` | 3.67 | 0.57 | 224 | 1 |
| `pvt_v2_b1` | 14.01 | 2.12 | 224 | 1 |
| `pvt_v2_b2` | 25.36 | 4.05 | 224 | 1 |
| `pvt_v2_b2_li` | 22.55 | 3.91 | 224 | 1 |
| `pvt_v2_b3` | 45.24 | 6.92 | 224 | 1 |
| `pvt_v2_b4` | 62.56 | 10.14 | 224 | 1 |
| `pvt_v2_b5` | 81.96 | 11.76 | 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 |
|---|---:|---:|---:|---:|---|---|
| `pvt_v2_b0.in1k` | 224 | 70.654 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/pvt_v2_b0.in1k) |
| `pvt_v2_b1.in1k` | 224 | 78.698 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/pvt_v2_b1.in1k) |
| `pvt_v2_b2.in1k` | 224 | 82.062 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/pvt_v2_b2.in1k) |
| `pvt_v2_b2_li.in1k` | 224 | 82.200 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/pvt_v2_b2_li.in1k) |
| `pvt_v2_b3.in1k` | 224 | 83.120 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/pvt_v2_b3.in1k) |
| `pvt_v2_b4.in1k` | 224 | 83.710 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/pvt_v2_b4.in1k) |
| `pvt_v2_b5.in1k` | 224 | 83.760 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/pvt_v2_b5.in1k) |

## Citation

```bibtex
@misc{wang2021pvtv2,
      title={PVTv2: Improved Baselines with Pyramid Vision Transformer},
      author={Wenhai Wang and Enze Xie and Xiang Li and Deng-Ping Fan and Kaitao Song and Ding Liang and
        Tong Lu and Ping Luo and Ling Shao},
      year={2021},
      eprint={2106.13797},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}
```