Update README.md
Browse files
README.md
CHANGED
@@ -83,3 +83,26 @@ The following hyperparameters were used during training:
|
|
83 |
- Pytorch 1.13.1+cu117
|
84 |
- Datasets 2.9.0
|
85 |
- Tokenizers 0.13.2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
83 |
- Pytorch 1.13.1+cu117
|
84 |
- Datasets 2.9.0
|
85 |
- Tokenizers 0.13.2
|
86 |
+
|
87 |
+
### Code to Run
|
88 |
+
def vit_classify(image):
|
89 |
+
vit = ViTForImageClassification.from_pretrained("oschamp/vit-artworkclassifier")
|
90 |
+
vit.eval()
|
91 |
+
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
92 |
+
vit.to(device)
|
93 |
+
|
94 |
+
model_name_or_path = 'google/vit-base-patch16-224-in21k'
|
95 |
+
feature_extractor = ViTFeatureExtractor.from_pretrained(model_name_or_path)
|
96 |
+
|
97 |
+
#LOAD IMAGE
|
98 |
+
|
99 |
+
encoding = feature_extractor(images=image, return_tensors="pt")
|
100 |
+
encoding.keys()
|
101 |
+
|
102 |
+
pixel_values = encoding['pixel_values'].to(device)
|
103 |
+
|
104 |
+
outputs = vit(pixel_values)
|
105 |
+
logits = outputs.logits
|
106 |
+
|
107 |
+
prediction = logits.argmax(-1)
|
108 |
+
return prediction.item() #vit.config.id2label[prediction.item()]
|