add requirements.txt

app.py

@@ -1,8 +1,7 @@
+import clip
 import gradio as gr
 import os
 import torch
-from torchvision import transforms
-from PIL import Image
 from torchvision.datasets import CIFAR100
 from transformers import CLIPProcessor, CLIPModel
 
@@ -11,32 +10,17 @@ processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
 
 cifar100 = CIFAR100(root=os.path.expanduser("~/.cache"), download=True, train=False)
 
-IMG_SIZE = 32 if torch.cuda.is_available() else 32
-COMPOSED_TRANSFORMERS = transforms.Compose([
-    transforms.Resize(IMG_SIZE),
-    transforms.ToTensor(),
-])
+text_inputs = torch.cat([clip.tokenize(f"a photo of a {c}") for c in cifar100.classes])
 
-NORMALIZE_TENSOR = transforms.Normalize(
-    mean=[0.485, 0.456, 0.406],
-    std=[0.229, 0.224, 0.225]
-)
 
+# TODO debug cette ligne pour avoir un affichage correct
 
-def np_array_to_tensor_image(img, width=IMG_SIZE, height=IMG_SIZE, device='cpu'):
-    image = Image.fromarray(img).convert('RGB').resize((width, height))
-    image = COMPOSED_TRANSFORMERS(image).unsqueeze(0)
-    return image.to(device, torch.float)
 
-
-def normalize_tensor(tensor: torch.tensor) -> torch.tensor:
-    return NORMALIZE_TENSOR(tensor)
+# TODO Finir l'affichage du résultat
 
 
 def send_inputs(img):
-    ##img = np_array_to_tensor_image(img)
-    ##img = normalize_tensor(img)
-    inputs = processor(images=img, return_tensors="pt", padding=True)
+    inputs = processor(text=cifar100.classes, images=img, return_tensors="pt", padding=True)
     outputs = model(**inputs)
     logits_per_image = outputs.logits_per_image
     probs = logits_per_image.softmax(dim=1)

requirements.txt

@@ -2,4 +2,5 @@ torch~=1.11.0
 torchvision~=0.12.0
 gradio~=3.0.2
 Pillow~=9.0.1
-transformers~=4.19.4
+transformers~=4.19.4
+clip~=0.2.0