README.md

---
tags:
- image-classification
- timm
- transformers
- animetimm
- dghs-imgutils
library_name: timm
license: gpl-3.0
datasets:
- animetimm/danbooru-wdtagger-v4-w640-ws-full
base_model:
- timm/convnext_base.fb_in22k_ft_in1k
---

# Anime Tagger convnext_base.dbv4-full

## Model Details

- **Model Type:** Multilabel Image classification / feature backbone
- **Model Stats:**
  - Params: 100.4M
  - FLOPs / MACs: 123.1G / 61.4G
  - Image size: train = 448 x 448, test = 448 x 448
- **Dataset:** [animetimm/danbooru-wdtagger-v4-w640-ws-full](https://huggingface.co/datasets/animetimm/danbooru-wdtagger-v4-w640-ws-full)
  - Tags Count: 12476
    - General (#0) Tags Count: 9225
    - Character (#4) Tags Count: 3247
    - Rating (#9) Tags Count: 4

## Results

|     #      |    Macro@0.40 (F1/MCC/P/R)    |    Micro@0.40 (F1/MCC/P/R)    |  Macro@Best (F1/P/R)  |
|:----------:|:-----------------------------:|:-----------------------------:|:---------------------:|
| Validation | 0.521 / 0.525 / 0.538 / 0.524 | 0.669 / 0.668 / 0.663 / 0.675 |          ---          |
|    Test    | 0.521 / 0.525 / 0.538 / 0.525 | 0.669 / 0.668 / 0.663 / 0.675 | 0.553 / 0.564 / 0.565 |

* `Macro/Micro@0.40` means the metrics on the threshold 0.40.
* `Macro@Best` means the mean metrics on the tag-level thresholds on each tags, which should have the best F1 scores.

## Thresholds

|  Category  |   Name    |  Alpha  |  Threshold  |  Micro@Thr (F1/P/R)   |  Macro@0.40 (F1/P/R)  |  Macro@Best (F1/P/R)  |
|:----------:|:---------:|:-------:|:-----------:|:---------------------:|:---------------------:|:---------------------:|
|     0      |  general  |    1    |     0.4     | 0.657 / 0.652 / 0.662 | 0.398 / 0.426 / 0.397 | 0.431 / 0.434 / 0.457 |
|     4      | character |    1    |    0.66     | 0.907 / 0.939 / 0.877 | 0.870 / 0.857 / 0.887 | 0.900 / 0.933 / 0.873 |
|     9      |  rating   |    1    |     0.4     | 0.823 / 0.784 / 0.866 | 0.829 / 0.801 / 0.860 | 0.830 / 0.806 / 0.857 |

* `Micro@Thr` means the metrics on the category-level suggested thresholds, which are listed in the table above.
* `Macro@0.40` means the metrics on the threshold 0.40.
* `Macro@Best` means the metrics on the tag-level thresholds on each tags, which should have the best F1 scores.

For tag-level thresholds, you can find them in [selected_tags.csv](https://huggingface.co/animetimm/convnext_base.dbv4-full/resolve/main/selected_tags.csv).

## How to Use

We provided a sample image for our code samples, you can find it [here](https://huggingface.co/animetimm/convnext_base.dbv4-full/blob/main/sample.webp).

### Use TIMM And Torch

Install [dghs-imgutils](https://github.com/deepghs/imgutils), [timm](https://github.com/huggingface/pytorch-image-models) and other necessary requirements with the following command

```shell
pip install 'dghs-imgutils>=0.17.0' torch huggingface_hub timm pillow pandas
```

After that you can load this model with timm library, and use it for train, validation and test, with the following code

```python
import json

import pandas as pd
import torch
from huggingface_hub import hf_hub_download
from imgutils.data import load_image
from imgutils.preprocess import create_torchvision_transforms
from timm import create_model

repo_id = 'animetimm/convnext_base.dbv4-full'
model = create_model(f'hf-hub:{repo_id}', pretrained=True)
model.eval()

with open(hf_hub_download(repo_id=repo_id, repo_type='model', filename='preprocess.json'), 'r') as f:
    preprocessor = create_torchvision_transforms(json.load(f)['test'])
# Compose(
#     PadToSize(size=(512, 512), interpolation=bilinear, background_color=white)
#     Resize(size=448, interpolation=bicubic, max_size=None, antialias=True)
#     CenterCrop(size=[448, 448])
#     MaybeToTensor()
#     Normalize(mean=tensor([0.4850, 0.4560, 0.4060]), std=tensor([0.2290, 0.2240, 0.2250]))
# )

image = load_image('https://huggingface.co/animetimm/convnext_base.dbv4-full/resolve/main/sample.webp')
input_ = preprocessor(image).unsqueeze(0)
# input_, shape: torch.Size([1, 3, 448, 448]), dtype: torch.float32
with torch.no_grad():
    output = model(input_)
    prediction = torch.sigmoid(output)[0]
# output, shape: torch.Size([1, 12476]), dtype: torch.float32
# prediction, shape: torch.Size([12476]), dtype: torch.float32

df_tags = pd.read_csv(
    hf_hub_download(repo_id=repo_id, repo_type='model', filename='selected_tags.csv'),
    keep_default_na=False
)
tags = df_tags['name']
mask = prediction.numpy() >= df_tags['best_threshold']
print(dict(zip(tags[mask].tolist(), prediction[mask].tolist())))
# {'sensitive': 0.8121482729911804,
#  '1girl': 0.9985615611076355,
#  'solo': 0.9850295186042786,
#  'looking_at_viewer': 0.7997868061065674,
#  'blush': 0.8888278603553772,
#  'smile': 0.9216164350509644,
#  'short_hair': 0.7902982234954834,
#  'shirt': 0.4786570370197296,
#  'long_sleeves': 0.7071284055709839,
#  'brown_hair': 0.5399358868598938,
#  'holding': 0.6793648600578308,
#  'dress': 0.6637858152389526,
#  'closed_mouth': 0.3529078960418701,
#  'sitting': 0.6096656322479248,
#  'purple_eyes': 0.9120762348175049,
#  'flower': 0.9145119786262512,
#  'braid': 0.8637544512748718,
#  'red_hair': 0.5444879531860352,
#  'blunt_bangs': 0.3928454518318176,
#  'tears': 0.7742242217063904,
#  'plant': 0.8320311307907104,
#  'blue_flower': 0.4381242096424103,
#  'crown_braid': 0.7148879766464233,
#  'potted_plant': 0.8044518232345581,
#  'flower_pot': 0.7585387229919434,
#  'pavement': 0.21422843635082245,
#  'wiping_tears': 0.8426964282989502,
#  'stone_floor': 0.09097751975059509,
#  'cobblestone': 0.12426003068685532}
```
### Use ONNX Model For Inference

Install [dghs-imgutils](https://github.com/deepghs/imgutils) with the following command

```shell
pip install 'dghs-imgutils>=0.17.0'
```

Use `multilabel_timm_predict` function with the following code

```python
from imgutils.generic import multilabel_timm_predict

general, character, rating = multilabel_timm_predict(
    'https://huggingface.co/animetimm/convnext_base.dbv4-full/resolve/main/sample.webp',
    repo_id='animetimm/convnext_base.dbv4-full',
    fmt=('general', 'character', 'rating'),
)

print(general)
# {'1girl': 0.9985615611076355,
#  'solo': 0.9850296378135681,
#  'smile': 0.9216161966323853,
#  'flower': 0.9145116806030273,
#  'purple_eyes': 0.9120762944221497,
#  'blush': 0.8888277411460876,
#  'braid': 0.863754391670227,
#  'wiping_tears': 0.8426950573921204,
#  'plant': 0.8320306539535522,
#  'potted_plant': 0.8044508695602417,
#  'looking_at_viewer': 0.7997866868972778,
#  'short_hair': 0.7902982234954834,
#  'tears': 0.7742242813110352,
#  'flower_pot': 0.758537232875824,
#  'crown_braid': 0.7148873209953308,
#  'long_sleeves': 0.7071276307106018,
#  'holding': 0.6793639659881592,
#  'dress': 0.6637860536575317,
#  'sitting': 0.6096659898757935,
#  'red_hair': 0.544488251209259,
#  'brown_hair': 0.539935827255249,
#  'shirt': 0.4786553978919983,
#  'blue_flower': 0.4381209909915924,
#  'blunt_bangs': 0.39284440875053406,
#  'closed_mouth': 0.35290712118148804,
#  'pavement': 0.21422752737998962,
#  'cobblestone': 0.12425824999809265,
#  'stone_floor': 0.0909762978553772}
print(character)
# {}
print(rating)
# {'sensitive': 0.812149167060852}
```

For further information, see [documentation of function multilabel_timm_predict](https://dghs-imgutils.deepghs.org/main/api_doc/generic/multilabel_timm.html#multilabel-timm-predict).