first version of ffmpeg command

app.py

@@ -1,8 +1,9 @@
 import os
+import subprocess
 
 import streamlit as st
 
-from utils import get_configs, get_display_names, get_path_for_viz
+from utils import get_configs, get_display_names, get_path_for_viz, Layout
 
 # st.header("EVREAL  - Event-based Video Reconstruction Evaluation and Analysis Library")
 #
@@ -20,6 +21,8 @@ st.title("Result Analysis Tool")
 
 data_base_path = "/home/bercan/ebv/evreal_data"
 
+font_path = "font/Ubuntu-B.ttf"
+
 dataset_cfg_path = os.path.join("cfg", "dataset")
 model_cfg_path = os.path.join("cfg", "model")
 metric_cfg_path = os.path.join("cfg", "metric")
@@ -75,10 +78,10 @@ model_only_viz = [viz for viz in selected_visualizations if viz['viz_type'] == '
 both_viz = [viz for viz in selected_visualizations if viz['viz_type'] == 'both']
 
 recon_viz = {"name": "recon", "display_name": "Reconstruction", "viz_type": "both", "gt_type": "frame"}
-ground_truth = {"name": "gt", "display_name": "Ground Truth", "model_id": "groundtruth"}
+# ground_truth = {"name": "gt", "display_name": "Ground Truth", "model_id": "groundtruth"}
 
 model_viz = [recon_viz] + both_viz + selected_metrics + model_only_viz
-num_model_rows = len(model_viz) + 1
+num_model_rows = len(model_viz)
 
 gt_viz = []
 if selected_dataset['has_frames']:
@@ -87,28 +90,55 @@ if selected_dataset['has_frames']:
     gt_viz.extend([viz for viz in gt_only_viz if viz['gt_type'] == 'frame'])
 
 gt_viz.extend([viz for viz in both_viz if viz['gt_type'] == 'event'])
-gt_viz.extend([viz for viz in gt_viz if viz['gt_type'] == 'event'])
+gt_viz.extend([viz for viz in gt_only_viz if viz['gt_type'] == 'event'])
+
+# print(get_display_names(model_viz))
+# print(get_display_names(gt_viz))
+# st.stop()
 
-num_gt_rows = len(gt_viz) + 1
+num_gt_rows = len(gt_viz)
 num_rows = max(num_model_rows, num_gt_rows)
 
-num_model_columns = len(selected_models) + 1
+num_model_columns = len(selected_models)
+
+num_elements = num_rows * num_model_columns
 
+layout = Layout(num_rows, num_model_columns)
+layout_str = layout.get_layout_str()
+
+video_paths = []
 for row_idx in range(num_rows):
-    row_visualizations = []
     for col_idx in range(num_model_columns):
-        if row_idx == 0 and col_idx == 0:
-            print("meta")
-            pass
-        elif row_idx == 0:
-            # model names
-            print(selected_models[col_idx - 1]['display_name'])
-            pass
-        elif col_idx == 0:
-            # metric names
-            print(model_viz[row_idx - 1]['display_name'])
-            pass
-        else:
-            video_path = get_path_for_viz(data_base_path, selected_dataset, selected_sequence, selected_models[col_idx - 1], model_viz[row_idx - 1])
-            print(video_path)
-            pass
+        video_path = get_path_for_viz(data_base_path, selected_dataset, selected_sequence,
+                                      selected_models[col_idx], model_viz[row_idx])
+        print(video_path)
+        video_paths.append(video_path)
+#         if os.path.isfile(video_path):
+#             video_paths.append(video_path)
+#         else:
+#             print("Video path does not exist: " + video_path)
+#
+# assert len(video_paths) == num_elements, "Number of video paths is not equal to expected number of elements"
+
+inputs_str = " ".join(["-i " + video_path for video_path in video_paths])
+
+crop_str = "crop=trunc(iw/2)*2:trunc(ih/2)*2"
+
+w = selected_dataset["width"]
+input_scaling_parts = []
+xstack_input_parts = []
+for i in range(num_elements):
+    input_scaling_part = "[{}:v]scale={}:-1,{}[v{}]".format(i, w, crop_str, i)
+    input_scaling_parts.append(input_scaling_part)
+    xstack_input_part = "[v{}]".format(i)
+    xstack_input_parts.append(xstack_input_part)
+input_scaling_str = ";".join(input_scaling_parts)
+xstack_input_str = "".join(xstack_input_parts)
+
+
+
+# opt = "-c:v libx264 -preset veryslow -crf 18 -c:a copy"
+opt = ""
+ffmpeg_command_str = "ffmpeg -y " + inputs_str + " -filter_complex \"" + input_scaling_str + ";" + xstack_input_str + "xstack=inputs=" + str(num_elements) + ":layout=" + layout_str + "\"" + opt + " output.mp4"
+print(ffmpeg_command_str)
+subprocess.call(ffmpeg_command_str, shell=True)

utils.py

@@ -36,3 +36,38 @@ def get_path_for_viz(base_path, dataset, sequence, model, viz):
     return video_path
 
 
+def get_text_video(video_h, video_w, text, font, fontsize, fontcolor="black"):
+    pass
+
+
+class Layout:
+    def __init__(self, num_rows, num_cols):
+        self.num_rows = num_rows
+        self.num_cols = num_cols
+
+    def get_layout_str(self):
+        layout_parts = []
+        layout_y_parts = []
+        for row_idx in range(self.num_rows):
+            layout_x_parts = []
+            if row_idx == 0:
+                layout_y_parts.append("0")
+            else:
+                upper_vid_no = (row_idx - 1) * self.num_cols
+                if row_idx > 1:
+                    layout_y_parts.append(layout_y_parts[-1] + "+h" + str(upper_vid_no))
+                else:
+                    layout_y_parts.append("h" + str(upper_vid_no))
+            for col_idx in range(self.num_cols):
+                if col_idx == 0:
+                    layout_x_parts.append("0")
+                else:
+                    left_vid_no = row_idx * self.num_cols + col_idx - 1
+                    if col_idx > 1:
+                        layout_x_parts.append(layout_x_parts[-1] + "+w" + str(left_vid_no))
+                    else:
+                        layout_x_parts.append("w" + str(left_vid_no))
+                layout_part = "_".join([layout_x_parts[-1], layout_y_parts[-1]])
+                layout_parts.append(layout_part)
+        layout_string = "|".join(layout_parts)
+        return layout_string