Create app.py

app.py

@@ -0,0 +1,56 @@
+import gradio as gr
+import numpy as np
+import pandas as pd
+import torch
+from datasets import load_dataset
+from transformers import CLIPProcessor, CLIPModel
+
+# Load dataset
+ds = load_dataset("amaye15/landscapes")
+train_ds = ds["train"]
+
+# Load embeddings
+df = pd.read_parquet("image_embeddings_clip.parquet")
+image_indices = df["image_index"].values
+emb_matrix = df.drop(columns=["image_index"]).values.astype(np.float32)
+
+# Load CLIP
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model_name = "openai/clip-vit-base-patch32"
+processor = CLIPProcessor.from_pretrained(model_name)
+model = CLIPModel.from_pretrained(model_name).to(device)
+model.eval()
+
+def l2_normalize(x):
+    return x / np.linalg.norm(x)
+
+@torch.no_grad()
+def embed_image(img):
+    inputs = processor(images=img, return_tensors="pt")
+    inputs = {k: v.to(device) for k, v in inputs.items()}
+    feats = model.get_image_features(**inputs)
+    feats = feats / feats.norm(dim=-1, keepdim=True)
+    return feats.squeeze(0).cpu().numpy()
+
+def recommend(img):
+    q_emb = embed_image(img)
+    sims = emb_matrix @ l2_normalize(q_emb)
+    top = np.argsort(-sims)[1:4]
+    results = []
+    for i in top:
+        results.append(train_ds[int(image_indices[i])]["pixel_values"])
+    return results
+
+demo = gr.Interface(
+    fn=recommend,
+    inputs=gr.Image(type="pil", label="Upload a landscape image"),
+    outputs=[
+        gr.Image(label="Recommendation 1"),
+        gr.Image(label="Recommendation 2"),
+        gr.Image(label="Recommendation 3"),
+    ],
+    title="Landscape Image Recommendation System",
+    description="Upload a landscape image and receive visually similar recommendations."
+)
+
+demo.launch()