Tidy up

app.py

@@ -127,38 +127,31 @@ EXAMPLES = [
     ["Intel/tvp-base", "a person reads a book.", "./examples/book.mp4", ],
 ]
 
-model_checkpoint = gr.Dropdown(MODELS, label="Model", value=MODELS[0], type="value")
-video_in = gr.Video(label="Video File", elem_id="video_in")
-# text_in = gr.Textbox(label="Text", placeholder="Description of event in the video", interactive=True)
-# text_out = gr.Textbox(label="Prediction", placeholder="Predicted start and end time")
-# examples = gr.Examples(examples=EXAMPLES, fn=predict_durations, inputs=[model_checkpoint, text_in, video_in], outputs=[text_out], cache_examples=True, preprocess=False)
-
-
 title = "Video Grounding with TVP"
 DESCRIPTION = """# Video Grounding with TVP"""
-css = """.toast-wrap { display: none !important } """
+
 with gr.Blocks(title=title) as demo:
     gr.Markdown(DESCRIPTION)
+    gr.Markdown(
+        """
+        Video Grounding is the task of localizing a moment in a video that best matches a natural language description.
+        For example, given the video of a person sitting on a bed, the model should be able to predict the start and end time of the video that best matches the description "a person is sitting on a bed".
+        Enter a description of an event in the video and select a video to see the predicted start and end time.
+        """
+    )
     with gr.Row():
         model_checkpoint = gr.Dropdown(MODELS, label="Model", value=MODELS[0], type="value")
-        # model_checkpoint.render()
 
-    with gr.Row():
-        with gr.Column():
-            video_in.render()
+    with gr.Row(equal_height=True):
+        with gr.Column(scale=0.5):
+            video_in = gr.Video(label="Video File", elem_id="video_in")
 
         with gr.Column():
             text_in = gr.Textbox(label="Text", placeholder="Description of event in the video", interactive=True)
             text_out = gr.Textbox(label="Prediction", placeholder="Predicted start and end time")
-            # text_in #.render()
             time_button = gr.Button("Get start and end time")
             time_button.click(predict_durations, inputs=[model_checkpoint, text_in, video_in], outputs=[text_out])
-            # text_out #.render()
-
-    with gr.Row():
-        examples = gr.Examples(examples=EXAMPLES, fn=predict_durations, inputs=[model_checkpoint, text_in, video_in], outputs=[text_out], cache_examples=True, preprocess=False)
-        # examples.render()
-        # text_out.render()
+            examples = gr.Examples(examples=EXAMPLES, fn=predict_durations, inputs=[model_checkpoint, text_in, video_in], outputs=[text_out], cache_examples=True, preprocess=False)
 
 
 if __name__ == "__main__":

gradio_cached_examples/12/log.csv

@@ -0,0 +1,4 @@
+Prediction,flag,username,timestamp
+"start: 0.0s, end: 6.8s",,,2023-11-22 16:36:10.654439
+"start: 0.0s, end: 11.4s",,,2023-11-22 16:36:26.574274
+"start: 0.0s, end: 5.6s",,,2023-11-22 16:36:42.339253

gradio_cached_examples/13/log.csv

@@ -0,0 +1,4 @@
+Prediction,flag,username,timestamp
+"start: 0.0s, end: 6.8s",,,2023-11-22 19:02:22.118880
+"start: 0.0s, end: 11.4s",,,2023-11-22 19:02:38.302300
+"start: 0.0s, end: 5.6s",,,2023-11-22 19:02:55.695928

gradio_cached_examples/14/log.csv

@@ -0,0 +1,4 @@
+Prediction,flag,username,timestamp
+"start: 0.0s, end: 6.8s",,,2023-11-22 16:26:11.313896
+"start: 0.0s, end: 11.4s",,,2023-11-22 16:26:27.749260
+"start: 0.0s, end: 5.6s",,,2023-11-22 16:26:43.506458