File size: 2,443 Bytes
568f229
2a70f66
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2299fd2
568f229
2a70f66
 
 
 
2299fd2
2a70f66
2299fd2
 
2a70f66
 
 
 
 
2299fd2
 
 
 
 
 
 
 
2a70f66
 
 
 
 
 
2299fd2
2a70f66
 
 
 
 
 
2299fd2
2a70f66
 
2299fd2
2a70f66
 
2299fd2
2a70f66
 
 
2299fd2
 
 
 
 
 
 
 
2a70f66
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
---
license: apache-2.0
tags:
- image-classification
- vision
- generated_from_trainer
datasets:
- imagefolder
metrics:
- accuracy
model-index:
- name: outputs
  results:
  - task:
      name: Image Classification
      type: image-classification
    dataset:
      name: imagefolder
      type: imagefolder
      config: default
      split: train
      args: default
    metrics:
    - name: Accuracy
      type: accuracy
      value: 0.7777777777777778
---


# Cowboy Hat emoji 🤠 (Western)

This model is a fine-tuned version of [facebook/convnextv2-large-22k-384](https://huggingface.co/facebook/convnextv2-large-22k-384) on the imagefolder dataset.
It achieves the following results on the evaluation set:
- Loss: 0.4250
- Accuracy: 0.7778

## Model description

When you want to know if an art is 🤠 or not 🤠.

- Current iteration: v3.5 (Continuous Image Integration)

## Wait, why?

gelbooru contains a lot of images, however not all of them are in the same region as south eas asia. As such, to filter out such images we have created a classifier that in theory learns the differences between western (USA/Europe/etc.) and not western (Japan/China/SEA).

The definition of "Not Western" is limited to the the asian region (Japan, Korea, China, Taiwan, Thailand and the surroundign region). The author believes that the art is similar enough with the same "style" which he personally prefers over western art.

## Intended uses & limitations

filter gelbooru data on 🤠 or not 🤠

## Training and evaluation data

Selected 358 images of 🤠 and not 🤠.

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 802565
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5.0

### Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
| 0.7384        | 1.0   | 152  | 0.4268          | 0.7963   |
| 0.2888        | 2.0   | 304  | 0.4250          | 0.7778   |
| 0.2953        | 3.0   | 456  | 0.4250          | 0.7778   |
| 0.4914        | 4.0   | 608  | 0.4250          | 0.7778   |
| 0.4099        | 5.0   | 760  | 0.4250          | 0.7778   |


### Framework versions

- Transformers 4.30.0.dev0
- Pytorch 1.13.1+cu117
- Datasets 2.12.0
- Tokenizers 0.13.3