add event type coloring

app.py

@@ -172,10 +172,11 @@ def inference(x, count_only_api, api_key, img_size=192, seq_len=64, stride_lengt
     periodLength = np.divide(period_lengths, period_length_overlaps, where=period_length_overlaps!=0)[:length]
     periodicity = np.divide(periodicities, period_length_overlaps, where=period_length_overlaps!=0)[:length]
     full_marks = np.divide(full_marks, period_length_overlaps, where=period_length_overlaps!=0)[:length]
-    event_type_logits = np.divide(event_type_logits, event_type_logit_overlaps, where=event_type_logit_overlaps!=0)[:length]
-    event_type_logits = np.mean(event_type_logits, axis=0)
+    per_frame_event_type_logits = np.divide(event_type_logits, event_type_logit_overlaps, where=event_type_logit_overlaps!=0)[:length]
+    event_type_logits = np.mean(per_frame_event_type_logits, axis=0)
     # softmax of event type logits  
     event_type_probs = np.exp(event_type_logits) / np.sum(np.exp(event_type_logits))
+    per_frame_event_types = np.argmax(per_frame_event_type_logits, axis=1)
     
     if median_pred_filter:
         periodicity = medfilt(periodicity, 5)
@@ -230,12 +231,14 @@ def inference(x, count_only_api, api_key, img_size=192, seq_len=64, stride_lengt
     jumps_per_second[misses] = 0
     frame_type = np.array(['miss' if miss else 'frame' for miss in misses])
     frame_type[full_marks > marks_threshold] = 'jump'
+    per_frame_event_types = per_frame_event_types / 6
     df = pd.DataFrame.from_dict({'period length': periodLength, 
                                  'jumping speed': jumping_speed,
                                 'jumps per second': jumps_per_second,
                                 'periodicity': periodicity,
                                 'miss': misses,
                                 'frame_type': frame_type,
+                                'event_type': per_frame_event_types,
                                 'jumps': full_marks,
                                 'jumps_size': (full_marks + 0.05) * 10,
                                 'miss_size': np.clip((1 - periodicity) * 0.9 + 0.1, 1, 10),
@@ -244,12 +247,12 @@ def inference(x, count_only_api, api_key, img_size=192, seq_len=64, stride_lengt
     fig = px.scatter(data_frame=df,
                     x='seconds', 
                     y='jumps per second',
-                    symbol='frame_type',
-                    symbol_map={'frame': 'circle', 'miss': 'circle-open', 'jump': 'triangle-down'},
-                    color='periodicity',
+                    #symbol='frame_type',
+                    #symbol_map={'frame': 'circle', 'miss': 'circle-open', 'jump': 'triangle-down'},
+                    color='event_type',
                     size='jumps_size',
                     size_max=10,
-                    color_continuous_scale='RdYlGn',
+                    color_continuous_scale='Rainbow',
                     title="Jumping speed (jumps-per-second)",
                     trendline='rolling',
                     trendline_options=dict(window=16),
@@ -302,7 +305,7 @@ with gr.Blocks(theme='WeixuanYuan/Soft_dark') as demo:
     gr.Markdown(DESCRIPTION)
     with gr.Column():
         with gr.Row():
-            in_video = gr.Video(label="Input Video", elem_id='input-video', format='mp4', width=400, height=400)
+            in_video = gr.Video(label="Input Video", elem_id='input-video', format='mp4', width=400, height=400, interactive=True)
             
     with gr.Row():
         run_button = gr.Button(label="Run", elem_id='run-button', scale=1)
@@ -356,6 +359,7 @@ with gr.Blocks(theme='WeixuanYuan/Soft_dark') as demo:
                         [os.path.join(os.path.dirname(__file__), "files", "train_95.mp4")],
                         [os.path.join(os.path.dirname(__file__), "files", "train_253.mp4")],
                         [os.path.join(os.path.dirname(__file__), "files", "train_66.mp4")],
+                        [os.path.join(os.path.dirname(__file__), "files", "train_21.mp4")]
                     ],
                 inputs=[in_video],
                 outputs=[out_text, out_plot, out_hist, out_event_type_dist],