Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,30 +1,35 @@
|
|
1 |
-
import torch
|
2 |
-
import re
|
3 |
import gradio as gr
|
4 |
-
from transformers import AutoTokenizer,ViTFeatureExtractor,VisionEncoderDecoderModel
|
5 |
|
|
|
6 |
device = 'cpu'
|
7 |
encoder_checkpoint = 'google/vit-base-patch16-224'
|
8 |
-
decoder_checkpoint = 'gpt2'
|
9 |
-
model_checkpoint = '
|
10 |
feature_extractor = ViTFeatureExtractor.from_pretrained(encoder_checkpoint)
|
|
|
11 |
model = VisionEncoderDecoderModel.from_pretrained(model_checkpoint).to(device)
|
12 |
|
13 |
|
14 |
-
|
|
|
15 |
image = image.convert('RGB')
|
16 |
-
image = feature_extractor(image,
|
17 |
-
clean_text = lambda x: x.replace('<|endoftext|>','').split('\n')[0]
|
18 |
caption_ids = model.generate(image, max_length = max_length)[0]
|
19 |
caption_text = clean_text(tokenizer.decode(caption_ids))
|
20 |
-
return caption_text
|
21 |
-
|
|
|
|
|
|
|
|
|
22 |
|
23 |
-
input = gr.inputs.Image(label='Image to generate caption',type = 'pil', optional=False)
|
24 |
-
output = gr.outputs.Textbox(type="auto",label="Caption")
|
25 |
|
26 |
-
article = "This
|
27 |
-
title = "Image Captioning"
|
28 |
|
29 |
interface = gr.Interface(
|
30 |
fn=predict,
|
|
|
1 |
+
import torch
|
2 |
+
import re
|
3 |
import gradio as gr
|
4 |
+
from transformers import AutoTokenizer, ViTFeatureExtractor, VisionEncoderDecoderModel
|
5 |
|
6 |
+
|
7 |
device = 'cpu'
|
8 |
encoder_checkpoint = 'google/vit-base-patch16-224'
|
9 |
+
decoder_checkpoint = 'surajp/gpt2-hindi'
|
10 |
+
model_checkpoint = 'team-indain-image-caption/hindi-image-captioning'
|
11 |
feature_extractor = ViTFeatureExtractor.from_pretrained(encoder_checkpoint)
|
12 |
+
tokenizer = AutoTokenizer.from_pretrained(decoder_checkpoint)
|
13 |
model = VisionEncoderDecoderModel.from_pretrained(model_checkpoint).to(device)
|
14 |
|
15 |
|
16 |
+
|
17 |
+
def predict(image,max_length=64, num_beams=4):
|
18 |
image = image.convert('RGB')
|
19 |
+
image = feature_extractor(image, return_tensors="pt").pixel_values.to(device)
|
20 |
+
clean_text = lambda x: x.replace('<|endoftext|>','').split('\n')[0]
|
21 |
caption_ids = model.generate(image, max_length = max_length)[0]
|
22 |
caption_text = clean_text(tokenizer.decode(caption_ids))
|
23 |
+
return caption_text
|
24 |
+
|
25 |
+
|
26 |
+
|
27 |
+
input = gr.inputs.Image(label="Image to search", type = 'pil', optional=False)
|
28 |
+
output = gr.outputs.Textbox(type="auto",label="Captions")
|
29 |
|
|
|
|
|
30 |
|
31 |
+
article = "This HuggingFace Space presents a demo for Image captioning in Hindi built with VIT Encoder and GPT2 Decoder"
|
32 |
+
title = "Hindi Image Captioning System"
|
33 |
|
34 |
interface = gr.Interface(
|
35 |
fn=predict,
|