|
--- |
|
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 |
|
|
