Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,17 +1,18 @@
|
|
1 |
import gradio as gr
|
2 |
from transformers import ViTImageProcessor, AutoModelForImageClassification
|
3 |
from PIL import Image
|
4 |
-
import io
|
5 |
import requests
|
6 |
-
from flask import Flask, request, jsonify
|
7 |
|
8 |
# Load the model and processor
|
9 |
processor = ViTImageProcessor.from_pretrained('AdamCodd/vit-base-nsfw-detector')
|
10 |
model = AutoModelForImageClassification.from_pretrained('AdamCodd/vit-base-nsfw-detector')
|
11 |
|
12 |
# Define prediction function
|
13 |
-
def predict_image(
|
14 |
try:
|
|
|
|
|
|
|
15 |
# Process the image and make prediction
|
16 |
inputs = processor(images=image, return_tensors="pt")
|
17 |
outputs = model(**inputs)
|
@@ -28,43 +29,10 @@ def predict_image(image):
|
|
28 |
# Create Gradio interface
|
29 |
iface = gr.Interface(
|
30 |
fn=predict_image,
|
31 |
-
inputs=gr.
|
32 |
outputs=gr.Textbox(label="Predicted Class"),
|
33 |
title="NSFW Image Classifier"
|
34 |
)
|
35 |
|
36 |
-
# Launch the
|
37 |
-
iface.launch()
|
38 |
-
|
39 |
-
# Flask app for API endpoint
|
40 |
-
app = Flask(__name__)
|
41 |
-
|
42 |
-
@app.route('/predict', methods=['POST'])
|
43 |
-
def predict():
|
44 |
-
if 'file' not in request.files:
|
45 |
-
return jsonify({'error': 'No file part'}), 400
|
46 |
-
|
47 |
-
file = request.files['file']
|
48 |
-
if file.filename == '':
|
49 |
-
return jsonify({'error': 'No selected file'}), 400
|
50 |
-
|
51 |
-
try:
|
52 |
-
# Load image from the uploaded file
|
53 |
-
image = Image.open(file.stream)
|
54 |
-
|
55 |
-
# Process the image and make prediction
|
56 |
-
inputs = processor(images=image, return_tensors="pt")
|
57 |
-
outputs = model(**inputs)
|
58 |
-
logits = outputs.logits
|
59 |
-
|
60 |
-
# Get predicted class
|
61 |
-
predicted_class_idx = logits.argmax(-1).item()
|
62 |
-
predicted_label = model.config.id2label[predicted_class_idx]
|
63 |
-
|
64 |
-
return jsonify({'predicted_class': predicted_label})
|
65 |
-
except Exception as e:
|
66 |
-
return jsonify({'error': str(e)}), 500
|
67 |
-
|
68 |
-
# Run Flask app
|
69 |
-
if __name__ == '__main__':
|
70 |
-
app.run(port=5000)
|
|
|
1 |
import gradio as gr
|
2 |
from transformers import ViTImageProcessor, AutoModelForImageClassification
|
3 |
from PIL import Image
|
|
|
4 |
import requests
|
|
|
5 |
|
6 |
# Load the model and processor
|
7 |
processor = ViTImageProcessor.from_pretrained('AdamCodd/vit-base-nsfw-detector')
|
8 |
model = AutoModelForImageClassification.from_pretrained('AdamCodd/vit-base-nsfw-detector')
|
9 |
|
10 |
# Define prediction function
|
11 |
+
def predict_image(image_url):
|
12 |
try:
|
13 |
+
# Load image from URL
|
14 |
+
image = Image.open(requests.get(image_url, stream=True).raw)
|
15 |
+
|
16 |
# Process the image and make prediction
|
17 |
inputs = processor(images=image, return_tensors="pt")
|
18 |
outputs = model(**inputs)
|
|
|
29 |
# Create Gradio interface
|
30 |
iface = gr.Interface(
|
31 |
fn=predict_image,
|
32 |
+
inputs=gr.Textbox(label="Image URL", placeholder="Enter image URL here"),
|
33 |
outputs=gr.Textbox(label="Predicted Class"),
|
34 |
title="NSFW Image Classifier"
|
35 |
)
|
36 |
|
37 |
+
# Launch the interface
|
38 |
+
iface.launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|