add more examples

app.py

@@ -35,12 +35,6 @@ model_ir = ie.read_model(model=model_xml)
 config = {"PERFORMANCE_HINT": "LATENCY"}
 compiled_model_ir = ie.compile_model(model=model_ir, device_name="CPU", config=config)
 
-a = os.path.join(os.path.dirname(__file__), "files", "dylan.mp4")
-b = os.path.join(os.path.dirname(__file__), "files", "train14.mp4")
-c = os.path.join(os.path.dirname(__file__), "files", "train_17.mp4")
-d = os.path.join(os.path.dirname(__file__), "files", "train13.mp4")
-e = os.path.join(os.path.dirname(__file__), "files", "train_213.mp4")
-
 
 class SquarePad:
     # https://discuss.pytorch.org/t/how-to-resize-and-pad-in-a-torchvision-transforms-compose/71850/9
@@ -243,9 +237,10 @@ def inference(x, count_only_api, api_key, img_size=192, seq_len=64, stride_lengt
                                 'miss': misses,
                                 'frame_type': frame_type,
                                 'jumps': full_marks,
-                                'jumps_size': (full_marks + 0.1) * 8,
+                                'jumps_size': (full_marks + 0.05) * 10,
                                 'miss_size': np.clip((1 - periodicity) * 0.9 + 0.1, 1, 10),
                                 'seconds': np.linspace(0, seconds, num=len(periodLength))})
+    
     fig = px.scatter(data_frame=df,
                     x='seconds', 
                     y='jumps per second',
@@ -261,6 +256,7 @@ def inference(x, count_only_api, api_key, img_size=192, seq_len=64, stride_lengt
                     trendline_color_override="goldenrod",
                     trendline_scope='overall',
                     template="plotly_dark")
+    
     fig.update_layout(legend=dict(
         orientation="h",
         yanchor="bottom",
@@ -332,30 +328,31 @@ with gr.Blocks(theme='WeixuanYuan/Soft_dark') as demo:
               
     with gr.Accordion(label="Instructions and more information", open=False):
         instructions = "## Instructions:"
-        instructions += "\n* Upload a video and click 'Run' to get a prediction of the number of jumps (either one foot, or both). This could take a couple minutes! The model is trained on single rope and double dutch speed, but try out any videos you want."
-        instructions += "\n* If you know the true count, you can provide it and your video will be used later to improve the model."
+        instructions += "\n* Upload a video and click 'Run' to get a prediction of the number of jumps (either one foot, or both). This could take a couple minutes!"
         instructions += "\n\n## Tips (optional):"
         instructions += "\n* Trim the video to start and end of the event"
         instructions += "\n* Frame the jumper fully, in the center of the frame"
         instructions += "\n* Videos are automatically resized, so higher resolution will not help, but a closer framing of the jumper might help. Try cropping the video differently."
-        instructions += "\n\n\nUnfortunately due to inference costs, right now we have to deploy a slower, less accurate version of the model but it still works surprisingly well in most cases. If you run into any issues let us know. If you would like to contribute to the project, please reach out!"
         gr.Markdown(instructions)
 
         faq = "## FAQ:"
         faq += "\n* **Q:** Does the model recognize misses?\n    * **A:** Yes, but if it fails, you can try tuning the miss threshold slider to make it more sensitive."
         faq += "\n* **Q:** Does the model recognize double dutch?\n    * **A:** Yes, but it is trained on a smaller set of double dutch videos, so it may not work perfectly."
-        faq += "\n* **Q:** Does the model recognize double unders\n    * **A:** Yes, but it is trained on a smaller set of double under videos, so it may not work perfectly. It is also trained to count the rope, not the jumps so you will need to divide the count by 2."
-        faq += "\n* **Q:** Does the model count both feet?\n    * **A:** Yes, but for convention we usually return the halved (one foot) score."
+        faq += "\n* **Q:** Does the model recognize double unders\n    * **A:** Yes, but it is trained on a smaller set of double under videos, so it may not work perfectly. It is also trained to count the rope, not the jumps so you will need to divide the count by 2 to get the traditional double under count."
+        faq += "\n* **Q:** Does the model count both feet?\n    * **A:** Yes, it counts every time the rope goes around no matter the event."
         gr.Markdown(faq)
 
     demo_inference = partial(inference, count_only_api=False, api_key=None)
 
     gr.Examples(examples=[
-                        [a],
-                        [b],
-                        [c],
-                        [d],
-                        [e]
+                        [os.path.join(os.path.dirname(__file__), "files", "dylan.mp4")],
+                        [os.path.join(os.path.dirname(__file__), "files", "train14.mp4")],
+                        [os.path.join(os.path.dirname(__file__), "files", "train_17.mp4")],
+                        [os.path.join(os.path.dirname(__file__), "files", "train13.mp4")],
+                        [os.path.join(os.path.dirname(__file__), "files", "train_213.mp4")],
+                        [os.path.join(os.path.dirname(__file__), "files", "train_156.mp4")],
+                        [os.path.join(os.path.dirname(__file__), "files", "train_202.mp4")],
+                        [os.path.join(os.path.dirname(__file__), "files", "train_57.mp4")],
                     ],
                 inputs=[in_video],
                 outputs=[out_text, out_plot, out_hist, out_event_type_dist],