Create app.py

app.py

@@ -0,0 +1,59 @@
+import gradio as gr
+from datasets import load_dataset, Image
+import random
+import numpy as np
+from PIL import Image
+from transformers import CLIPProcessor, CLIPModel
+
+model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
+processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
+
+# Load the Unsplash dataset
+dataset = load_dataset("jamescalam/unsplash-25k-photos", split="train")
+
+labels = []
+
+def predict(image):
+    inputs = processor(text=[f"a photo of {c}" for c in labels], images=image, return_tensors="pt", padding=True)
+    outputs = model(**inputs)
+    logits_per_image = outputs.logits_per_image # this is the image-text similarity score
+    probs = logits_per_image.softmax(dim=1) # we can take the softmax to get the label probabilities
+    return {k: float(v) for k, v in zip(labels, probs[0])}
+
+
+def rand_image():
+    n = dataset.num_rows
+    r = random.randrange(0,n)
+    return dataset[r]["photo_image_url"]+"?h=250"
+
+def set_labels(text):
+    global labels
+    labels = text.split(",")
+    
+
+with gr.Blocks() as demo:
+    instructions = """## Instructions:
+    1. Enter list of labels separated by commas (or select one of the examples below)
+    2. Click **Get Random Image** to grab random image from dataset and analyze it against the labels
+    3. Click **Re-Classify Image** to re-run classification on current image after changing labels"""
+    gr.Markdown(instructions)
+    with gr.Row(variant="compact"):
+        label_text = gr.Textbox(show_label=False, placeholder="enter labels").style(container=False)
+        #submit_btn = gr.Button("Submit").style(full_width=False)
+    gr.Examples(["spring, summer, fall, winter", "mountain, city, beach, ocean, desert, plains", "red, blue, green, white, black, purple, brown", "food, vehicle, none of the above"], inputs=label_text)
+    with gr.Row():
+        with gr.Column(variant="panel"):
+            im = gr.Image(interactive=False).style(height=250)
+            with gr.Row():
+                get_btn = gr.Button("Get Random Image").style(full_width=False)
+                reclass_btn = gr.Button("Re-Classify Image").style(full_width=False)
+        cf = gr.Label()
+    #submit_btn.click(fn=set_labels, inputs=label_text)
+    label_text.change(fn=set_labels, inputs=label_text)
+    label_text.blur(fn=set_labels, inputs=label_text)
+    label_text.submit(fn=set_labels, inputs=label_text)
+    get_btn.click(fn=rand_image, outputs=im)
+    im.change(predict, inputs=im, outputs=cf)
+    reclass_btn.click(predict, inputs=im, outputs=cf)
+
+demo.launch()