Spaces:
Sleeping
Sleeping
MUTED64
commited on
Commit
•
88b279a
1
Parent(s):
806212d
Add app and pth
Browse files- 1024_MLP_best-MSE4.1636_ep75.pth +3 -0
- app.py +36 -0
1024_MLP_best-MSE4.1636_ep75.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:617a1c3fe8cfdcfb79fa0df3d46d4673497ca47a8ee43ddd6d6a5027a478ec64
|
3 |
+
size 3716120
|
app.py
ADDED
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import torch
|
3 |
+
from PIL import Image
|
4 |
+
from torchvision.transforms import functional as F
|
5 |
+
from typing import List
|
6 |
+
from transformers import CLIPModel, CLIPProcessor
|
7 |
+
|
8 |
+
# Load the pre-trained model
|
9 |
+
model_path = "1024_MLP_best-MSE4.1636_ep75.pth"
|
10 |
+
model = torch.load(model_path)
|
11 |
+
model.eval()
|
12 |
+
|
13 |
+
# Load the CLIP model and processor
|
14 |
+
clip_model = CLIPModel.from_pretrained("ViT-L/14")
|
15 |
+
clip_processor = CLIPProcessor.from_pretrained("ViT-L/14")
|
16 |
+
|
17 |
+
# Define the prediction function
|
18 |
+
def predict(images: List[Image.Image]) -> float:
|
19 |
+
image_tensors = [F.to_tensor(img) for img in images]
|
20 |
+
inputs = clip_processor(images=image_tensors, return_tensors="pt", padding=True)
|
21 |
+
with torch.no_grad():
|
22 |
+
outputs = model(inputs.pixel_values)
|
23 |
+
scores = outputs.clamp(0, 10).cpu().numpy().reshape(-1).tolist()
|
24 |
+
return scores
|
25 |
+
|
26 |
+
# Define the Gradio interface
|
27 |
+
iface = gr.Interface(
|
28 |
+
fn=predict,
|
29 |
+
inputs="image",
|
30 |
+
outputs="number",
|
31 |
+
title="Kemono Aesthetic Scorer",
|
32 |
+
description="Predict the score of a kemono based on aesthetic features.",
|
33 |
+
)
|
34 |
+
|
35 |
+
# Run the Gradio interface
|
36 |
+
iface.launch()
|