# DLA

**Category:** CNN · **Code:**
[`timm/models/dla.py`](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/dla.py) ·
**Architectures:** 12 · **Pretrained weights:** 12 (12 with ImageNet-1k results)

Deep Layer Aggregation (DLA) networks merge features across blocks and stages with iterative and hierarchical
aggregation nodes. Includes DLA variants built from Res2Net blocks.

## Papers

* [Deep Layer Aggregation](https://arxiv.org/abs/1707.06484)
* [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('dla102x.in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `DLA` |
|---|---|
| [`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 (`DLA` with `dla46x_c` 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 |
|---|---:|---:|---:|---:|
| `dla34` | 15.74 | 3.07 | 224 | 1 |
| `dla46_c` | 1.30 | 0.58 | 224 | 1 |
| `dla46x_c` | 1.07 | 0.54 | 224 | 1 |
| `dla60` | 22.04 | 4.26 | 224 | 1 |
| `dla60_res2net` | 20.85 | 4.15 | 224 | 1 |
| `dla60_res2next` | 17.03 | 3.49 | 224 | 1 |
| `dla60x` | 17.35 | 3.54 | 224 | 1 |
| `dla60x_c` | 1.32 | 0.59 | 224 | 1 |
| `dla102` | 33.27 | 7.19 | 224 | 1 |
| `dla102x` | 26.31 | 5.89 | 224 | 1 |
| `dla102x2` | 41.28 | 9.34 | 224 | 1 |
| `dla169` | 53.39 | 11.60 | 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 |
|---|---:|---:|---:|---:|---|---|
| `dla34.in1k` | 224 | 74.632 | 224 | 1000 | bsd-3-clause | [timm](https://huggingface.co/timm/dla34.in1k) |
| `dla46_c.in1k` | 224 | 64.880 | 224 | 1000 | bsd-3-clause | [timm](https://huggingface.co/timm/dla46_c.in1k) |
| `dla46x_c.in1k` | 224 | 66.008 | 224 | 1000 | bsd-3-clause | [timm](https://huggingface.co/timm/dla46x_c.in1k) |
| `dla60.in1k` | 224 | 77.036 | 224 | 1000 | bsd-3-clause | [timm](https://huggingface.co/timm/dla60.in1k) |
| `dla60_res2net.in1k` | 224 | 78.480 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/dla60_res2net.in1k) |
| `dla60_res2next.in1k` | 224 | 78.488 | 224 | 1000 | unknown | [timm](https://huggingface.co/timm/dla60_res2next.in1k) |
| `dla60x.in1k` | 224 | 78.234 | 224 | 1000 | bsd-3-clause | [timm](https://huggingface.co/timm/dla60x.in1k) |
| `dla60x_c.in1k` | 224 | 67.928 | 224 | 1000 | bsd-3-clause | [timm](https://huggingface.co/timm/dla60x_c.in1k) |
| `dla102.in1k` | 224 | 78.016 | 224 | 1000 | bsd-3-clause | [timm](https://huggingface.co/timm/dla102.in1k) |
| `dla102x.in1k` | 224 | 78.516 | 224 | 1000 | bsd-3-clause | [timm](https://huggingface.co/timm/dla102x.in1k) |
| `dla102x2.in1k` | 224 | 79.448 | 224 | 1000 | bsd-3-clause | [timm](https://huggingface.co/timm/dla102x2.in1k) |
| `dla169.in1k` | 224 | 78.702 | 224 | 1000 | bsd-3-clause | [timm](https://huggingface.co/timm/dla169.in1k) |