Update app.py

app.py

@@ -31,6 +31,22 @@ VALID_ZEROSHOT_VIDEOCLASSIFICATION_MODELS = [
 processor = AutoProcessor.from_pretrained(DEFAULT_MODEL)
 model = AutoModel.from_pretrained(DEFAULT_MODEL)
 
+examples = [
+    [
+        "https://www.youtu.be/l1dBM8ZECao",
+        "sleeping dog,cat fight club,birds of prey",
+    ],
+    [
+        "https://youtu.be/VMj-3S1tku0",
+        "programming course,eating spaghetti,playing football",
+    ],
+    [
+        "https://youtu.be/BRw7rvLdGzU",
+        "game of thrones,the lord of the rings,vikings",
+    ],
+]
+
+
 def select_model(model_name):
     global processor, model
     processor = AutoProcessor.from_pretrained(model_name)
@@ -43,6 +59,8 @@ def predict(youtube_url_or_file_path, labels_text):
         video_path = download_youtube_video(youtube_url_or_file_path)
     else:
         video_path = youtube_url_or_file_path
+
+    # rearrange sampling rate based on video length and model input length
     num_total_frames = get_num_total_frames(video_path)
     num_model_input_frames = model.config.vision_config.num_frames
     if num_total_frames < FRAME_SAMPLING_RATE * num_model_input_frames:
@@ -60,6 +78,7 @@ def predict(youtube_url_or_file_path, labels_text):
     inputs = processor(
         text=labels, videos=list(frames), return_tensors="pt", padding=True
     )
+    # forward pass
     with torch.no_grad():
         outputs = model(**inputs)
 
@@ -74,7 +93,7 @@ def predict(youtube_url_or_file_path, labels_text):
 app = gr.Blocks()
 with app:
     gr.Markdown(
-        "# **<p align='center'> PROTOG - VIOLENCE DETECTION MODULE</p>**"
+        "# **<p align='center'>PROGTOG VIOLENCE DETECTION</p>**"
     )
 
     with gr.Row():
@@ -88,7 +107,7 @@ with app:
             model_names_dropdown.change(fn=select_model, inputs=model_names_dropdown)
             with gr.Tab(label="Youtube URL"):
                 gr.Markdown(
-                    "### **Enter Youtube URL**"
+                    "### **Youtube URL**"
                 )
                 youtube_url = gr.Textbox(label="Youtube URL:", show_label=True)
                 youtube_url_labels_text = gr.Textbox(
@@ -97,7 +116,7 @@ with app:
                 youtube_url_predict_btn = gr.Button(value="Predict")
             with gr.Tab(label="Local File"):
                 gr.Markdown(
-                    "### **Video Upload**"
+                    "### **Tags**"
                 )
                 video_file = gr.Video(label="Video File:", show_label=True)
                 local_video_labels_text = gr.Textbox(
@@ -112,6 +131,15 @@ with app:
         with gr.Column():
             predictions = gr.Label(label="Predictions:", show_label=True)
 
+    # gr.Markdown("**Examples:**")
+    # gr.Examples(
+    #     examples,
+    #     [youtube_url, youtube_url_labels_text],
+    #     [predictions, video_gif],
+    #     fn=predict,
+    #     cache_examples=True,
+    # )
+
     youtube_url_predict_btn.click(
         predict,
         inputs=[youtube_url, youtube_url_labels_text],
@@ -122,5 +150,11 @@ with app:
         inputs=[video_file, local_video_labels_text],
         outputs=[predictions, video_gif],
     )
+    # gr.Markdown(
+    #     """
+    #     \n Demo created by: <a href=\"https://github.com/fcakyon\">fcakyon</a>.
+    #     <br> Based on this <a href=\"https://huggingface.co/docs/transformers/main/model_doc/xclip">HuggingFace model</a>.
+    #     """
+    # )
 
 app.launch()