Update README.md
Browse files
README.md
CHANGED
|
@@ -15,24 +15,30 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 15 |
|
| 16 |
# Celebrity Classifier
|
| 17 |
## Model description
|
| 18 |
-
This model classifies a face to a celebrity. It is trained on [tonyassi/celebrity-1000](https://huggingface.co/datasets/tonyassi/celebrity-1000) and fine-tuned on [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k).
|
| 19 |
-
It achieves the following results on the evaluation set:
|
| 20 |
-
- Loss: 0.9089
|
| 21 |
-
- Accuracy: 0.7982
|
| 22 |
|
|
|
|
|
|
|
|
|
|
| 23 |
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
-
|
|
|
|
| 26 |
|
| 27 |
-
|
|
|
|
| 28 |
|
| 29 |
-
|
|
|
|
|
|
|
| 30 |
|
| 31 |
## Training and evaluation data
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
## Training procedure
|
| 36 |
|
| 37 |
### Training hyperparameters
|
| 38 |
|
|
@@ -48,32 +54,6 @@ The following hyperparameters were used during training:
|
|
| 48 |
- lr_scheduler_warmup_ratio: 0.1
|
| 49 |
- num_epochs: 20
|
| 50 |
|
| 51 |
-
### Training results
|
| 52 |
-
|
| 53 |
-
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|
| 54 |
-
|:-------------:|:-----:|:----:|:---------------:|:--------:|
|
| 55 |
-
| 0.2075 | 1.0 | 227 | 1.0255 | 0.7831 |
|
| 56 |
-
| 0.1359 | 2.0 | 455 | 1.1713 | 0.7517 |
|
| 57 |
-
| 0.1703 | 3.0 | 682 | 1.1582 | 0.7503 |
|
| 58 |
-
| 0.1052 | 4.0 | 910 | 1.1482 | 0.7567 |
|
| 59 |
-
| 0.0826 | 5.0 | 1137 | 1.1340 | 0.7514 |
|
| 60 |
-
| 0.1412 | 6.0 | 1365 | 1.1149 | 0.7514 |
|
| 61 |
-
| 0.105 | 7.0 | 1592 | 1.1071 | 0.7523 |
|
| 62 |
-
| 0.1067 | 8.0 | 1820 | 1.1161 | 0.7539 |
|
| 63 |
-
| 0.1329 | 9.0 | 2047 | 1.0587 | 0.7693 |
|
| 64 |
-
| 0.1196 | 10.0 | 2275 | 1.0416 | 0.7688 |
|
| 65 |
-
| 0.1368 | 11.0 | 2502 | 1.0618 | 0.7663 |
|
| 66 |
-
| 0.1162 | 12.0 | 2730 | 1.0285 | 0.7721 |
|
| 67 |
-
| 0.145 | 13.0 | 2957 | 1.0040 | 0.7776 |
|
| 68 |
-
| 0.1449 | 14.0 | 3185 | 0.9967 | 0.7800 |
|
| 69 |
-
| 0.1135 | 15.0 | 3412 | 0.9603 | 0.7842 |
|
| 70 |
-
| 0.1266 | 16.0 | 3640 | 0.9333 | 0.7861 |
|
| 71 |
-
| 0.1571 | 17.0 | 3867 | 0.9643 | 0.7836 |
|
| 72 |
-
| 0.278 | 18.0 | 4095 | 0.9526 | 0.7861 |
|
| 73 |
-
| 0.2596 | 19.0 | 4322 | 0.9022 | 0.7965 |
|
| 74 |
-
| 0.2432 | 19.96 | 4540 | 0.9089 | 0.7982 |
|
| 75 |
-
|
| 76 |
-
|
| 77 |
### Framework versions
|
| 78 |
|
| 79 |
- Transformers 4.35.2
|
|
|
|
| 15 |
|
| 16 |
# Celebrity Classifier
|
| 17 |
## Model description
|
| 18 |
+
This model classifies a face to a celebrity. It is trained on [tonyassi/celebrity-1000](https://huggingface.co/datasets/tonyassi/celebrity-1000) dataset and fine-tuned on [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k).
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
+
## Dataset description
|
| 21 |
+
[tonyassi/celebrity-1000](https://huggingface.co/datasets/tonyassi/celebrity-1000)
|
| 22 |
+
Top 1000 celebrities. 18,184 images. 256x256. Square cropped to face.
|
| 23 |
|
| 24 |
+
### How to use
|
| 25 |
+
```python
|
| 26 |
+
from transformers import pipeline
|
| 27 |
|
| 28 |
+
# Initialize image classification pipeline
|
| 29 |
+
pipe = pipeline("image-classification", model="tonyassi/celebrity-classifier")
|
| 30 |
|
| 31 |
+
# Perform classification
|
| 32 |
+
result = pipe('image.png')
|
| 33 |
|
| 34 |
+
# Print results
|
| 35 |
+
print(result)
|
| 36 |
+
```
|
| 37 |
|
| 38 |
## Training and evaluation data
|
| 39 |
+
It achieves the following results on the evaluation set:
|
| 40 |
+
- Loss: 0.9089
|
| 41 |
+
- Accuracy: 0.7982
|
|
|
|
| 42 |
|
| 43 |
### Training hyperparameters
|
| 44 |
|
|
|
|
| 54 |
- lr_scheduler_warmup_ratio: 0.1
|
| 55 |
- num_epochs: 20
|
| 56 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
### Framework versions
|
| 58 |
|
| 59 |
- Transformers 4.35.2
|