---
license: apache-2.0
tags:
- generated_from_trainer
metrics:
- precision
- recall
- f1
- accuracy
model-index:
- name: aesthetic_attribute_classifier
  results: []
widget:
- text: Check your vertical on the main support; it looks a little off. I'd also like to see how it looks with a bit of the sky cropped from the photo

---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# aesthetic_attribute_classifier

This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the [PCCD dataset](https://github.com/ivclab/DeepPhotoCritic-ICCV17).
It achieves the following results on the evaluation set:
- Loss: 0.3976
- Precision: {'precision': 0.877129341279301}
- Recall: {'recall': 0.8751381215469614}
- F1: {'f1': 0.875529982855803}
- Accuracy: {'accuracy': 0.8751381215469614}

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5

### Training results

| Training Loss | Epoch | Step | Validation Loss | Precision                         | Recall                         | F1                         | Accuracy                         |
|:-------------:|:-----:|:----:|:---------------:|:---------------------------------:|:------------------------------:|:--------------------------:|:--------------------------------:|
| 0.452         | 1.0   | 1528 | 0.4109          | {'precision': 0.8632779077963935} | {'recall': 0.8615101289134438} | {'f1': 0.8618616182904953} | {'accuracy': 0.8615101289134438} |
| 0.3099        | 2.0   | 3056 | 0.3976          | {'precision': 0.877129341279301}  | {'recall': 0.8751381215469614} | {'f1': 0.875529982855803}  | {'accuracy': 0.8751381215469614} |
| 0.227         | 3.0   | 4584 | 0.4320          | {'precision': 0.876211408446225}  | {'recall': 0.874401473296501}  | {'f1': 0.8747427955387239} | {'accuracy': 0.874401473296501}  |
| 0.1645        | 4.0   | 6112 | 0.4840          | {'precision': 0.8724641667216837} | {'recall': 0.8714548802946593} | {'f1': 0.8714577820909117} | {'accuracy': 0.8714548802946593} |
| 0.1141        | 5.0   | 7640 | 0.5083          | {'precision': 0.8755445355051571} | {'recall': 0.8747697974217311} | {'f1': 0.8748766125899489} | {'accuracy': 0.8747697974217311} |


### Framework versions

- Transformers 4.16.2
- Pytorch 1.10.2+cu113
- Datasets 1.18.3
- Tokenizers 0.11.0