Add application file

localfile.py

@@ -0,0 +1,41 @@
+import clip
+import torch
+import gradio as gr
+import torchvision.transforms as T
+from PIL import Image
+try:
+    from torchvision.transforms import InterpolationMode
+    BICUBIC = InterpolationMode.BICUBIC
+except ImportError:
+    BICUBIC = Image.BICUBIC
+import warnings
+warnings.filterwarnings("ignore")
+
+#MODEL_PATH = '/media/delta/S/clipmodel.pth'                        #CHANGE THIS IF YOU WANT TO CHANGE THE MODEL PATH
+MODEL_PATH ='/media/delta/S/clipmodel_large.pth'                    #CHANGE THIS IF YOU WANT TO CHANGE THE MODEL PATH
+
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model = clip.model.build_model(torch.load(MODEL_PATH)).to(device)
+preprocess = clip.clip._transform(model.visual.input_resolution)
+
+def zeroshot_detection(Press_Clear_Dont_Stack_Image):
+    inp = Press_Clear_Dont_Stack_Image
+
+    captions = "photo of a guardrail, no guardrail in the photo"   #CHANGE THIS IF YOU WANT TO CHANGE THE PREDICTION: separate by commas
+
+    captions = captions.split(',')
+    caption = clip.tokenize(captions).to(device)
+    image = preprocess(inp).unsqueeze(0).to(device)
+    with torch.no_grad():
+        image_features = model.encode_image(image)
+        text_features = model.encode_text(caption)
+    image_features /= image_features.norm(dim=-1, keepdim=True)
+    text_features /= text_features.norm(dim=-1, keepdim=True)
+    similarity = (100.0 * image_features @ text_features.T).softmax(dim=-1)
+    values, indices = similarity[0].topk(len(captions))
+    return {captions[indices[i].item()]: float(values[i].item()) for i in range(len(values))}   
+
+gr.Interface(fn=zeroshot_detection, 
+             inputs=[gr.Image(type="pil")],
+             outputs=gr.Label(num_top_classes=1)).launch()

requirements.txt

@@ -0,0 +1 @@
+git+https://github.com/openai/CLIP.git