Create app.py

app.py

@@ -0,0 +1,26 @@
+import gradio as gr
+import torch
+from transformers import VideoMAEForVideoClassification, VideoMAEFeatureExtractor
+from PIL import Image
+import numpy as np
+
+# Load model & processor
+model_name = "OPear/videomae-large-finetuned-UCF-Crime"
+model = VideoMAEForVideoClassification.from_pretrained(model_name)
+processor = VideoMAEFeatureExtractor.from_pretrained(model_name)
+
+def classify_video(video):
+    # video is a numpy array of shape (frames, H, W, C)
+    inputs = processor(video, return_tensors="pt")
+    with torch.no_grad():
+        outputs = model(**inputs)
+    logits = outputs.logits
+    predicted_class_idx = logits.argmax(-1).item()
+    return model.config.id2label[predicted_class_idx]
+
+iface = gr.Interface(fn=classify_video,
+                     inputs=gr.Video(),
+                     outputs="text",
+                     title="Video Classifier using VideoMAE")
+
+iface.launch()