Update README.md
Browse files
README.md
CHANGED
@@ -44,23 +44,45 @@ To use the model for inference:
|
|
44 |
from torchvision import transforms
|
45 |
import torch
|
46 |
from PIL import Image
|
|
|
47 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
|
|
48 |
# Load model
|
49 |
model = torch.load("EmoSet_clip_Lora_16.0R_8.0alphaLora_32_batch_0.0001_headmlp.pth").to(device).eval()
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
transform = transforms.Compose([
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
])
|
62 |
return transform
|
63 |
-
|
|
|
|
|
|
|
|
|
|
|
64 |
with torch.no_grad():
|
65 |
-
|
66 |
-
|
|
|
|
|
|
|
|
|
|
44 |
from torchvision import transforms
|
45 |
import torch
|
46 |
from PIL import Image
|
47 |
+
|
48 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
49 |
+
|
50 |
# Load model
|
51 |
model = torch.load("EmoSet_clip_Lora_16.0R_8.0alphaLora_32_batch_0.0001_headmlp.pth").to(device).eval()
|
52 |
+
|
53 |
+
# Emotion label mapping
|
54 |
+
idx2label = {
|
55 |
+
0: "amusement",
|
56 |
+
1: "awe",
|
57 |
+
2: "contentment",
|
58 |
+
3: "excitement",
|
59 |
+
4: "anger",
|
60 |
+
5: "disgust",
|
61 |
+
6: "fear",
|
62 |
+
7: "sadness"
|
63 |
+
}
|
64 |
+
|
65 |
+
# Preprocessing function
|
66 |
+
def emo_preprocess():
|
67 |
transform = transforms.Compose([
|
68 |
+
transforms.Resize(224),
|
69 |
+
transforms.CenterCrop(size=(224, 224)),
|
70 |
+
transforms.ToTensor(),
|
71 |
+
# Note: The model normalizes the image inside the forward pass
|
72 |
+
# using mean = (0.48145466, 0.4578275, 0.40821073) and
|
73 |
+
# std = (0.26862954, 0.26130258, 0.27577711)
|
74 |
+
])
|
75 |
return transform
|
76 |
+
|
77 |
+
# Load an image
|
78 |
+
image = Image.open("image_path.jpg").convert("RGB")
|
79 |
+
image = emo_preprocess()(image).unsqueeze(0).to(device)
|
80 |
+
|
81 |
+
# Run inference
|
82 |
with torch.no_grad():
|
83 |
+
outputs = model(image)
|
84 |
+
_, predicted = outputs.max(1) # Get the class index
|
85 |
+
|
86 |
+
# Get emotion label
|
87 |
+
predicted_emotion = idx2label[predicted.item()]
|
88 |
+
print(f"Predicted Emotion: {predicted_emotion}")
|