Spaces:
No application file
No application file
Rename image-app to image.py
Browse files
image-app
DELETED
|
File without changes
|
image.py
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 3 |
+
from PIL import Image
|
| 4 |
+
|
| 5 |
+
|
| 6 |
+
# Model and tokenizer loading
|
| 7 |
+
model_id = "vikhyatk/moondream2"
|
| 8 |
+
revision = "2024-03-06"
|
| 9 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 10 |
+
model_id, trust_remote_code=True, revision=revision
|
| 11 |
+
)
|
| 12 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id, revision=revision)
|
| 13 |
+
|
| 14 |
+
# Image loading
|
| 15 |
+
image_path = 'demp.png'
|
| 16 |
+
image = Image.open(image_path)
|
| 17 |
+
|
| 18 |
+
# Display the image
|
| 19 |
+
image.show()
|
| 20 |
+
|
| 21 |
+
# Encoding the image
|
| 22 |
+
enc_image = model.encode_image(image)
|
| 23 |
+
|
| 24 |
+
# Asking the model to describe the image
|
| 25 |
+
description = model.answer_question(enc_image, "Describe this image.", tokenizer)
|
| 26 |
+
print("Generated Description:", description)
|