second working version

app.py

@@ -3,7 +3,7 @@ import subprocess
 
 import streamlit as st
 
-from utils import get_configs, get_display_names, get_path_for_viz, Layout, get_video_height
+from utils import get_configs, get_display_names, get_path_for_viz, get_video_height
 
 # st.header("EVREAL  - Event-based Video Reconstruction Evaluation and Analysis Library")
 #
@@ -92,81 +92,67 @@ if selected_dataset['has_frames']:
 gt_viz.extend([viz for viz in both_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)
 num_rows = max(num_model_rows, num_gt_rows)
 
 if len(gt_viz) > 0:
     selected_models.append(ground_truth)
 
+padding = 2
 num_cols = len(selected_models)
-
+crop_str = "crop=trunc(iw/2)*2:trunc(ih/2)*2"
+pad_str = "pad=ceil(iw/2)*2+{}:ceil(ih/2)*2+{}:{}:{}".format(padding*2, padding*2, padding, padding)
 num_elements = num_rows * num_cols
 
+w = selected_dataset["width"]
+input_filter_parts = []
+xstack_input_parts = []
+layout_parts = []
 video_paths = []
-row_heights = [0]*num_rows
+row_heights = [""]*num_rows
 for row_idx in range(num_rows):
     for col_idx in range(num_cols):
         cur_model = selected_models[col_idx]
         if cur_model['name'] == "gt":
-            if row_idx >= len(gt_viz):
-                video_path = ""
-            else:
+            if row_idx < len(gt_viz):
                 video_path = get_path_for_viz(data_base_path, selected_dataset, selected_sequence, cur_model, gt_viz[row_idx])
                 # if not os.path.isfile(video_path):
                 #     raise ValueError("Video path does not exist: " + video_path)
-        else:
-            if row_idx >= len(model_viz):
-                video_path = ""
             else:
+                continue
+        else:
+            if row_idx < len(model_viz):
                 video_path = get_path_for_viz(data_base_path, selected_dataset, selected_sequence, cur_model, model_viz[row_idx])
                 # if not os.path.isfile(video_path):
                 #     raise ValueError("Video path does not exist: " + video_path)
-
-        if video_path and row_heights[row_idx] == 0:
-            row_heights[row_idx] = get_video_height(video_path)
-
-        print(video_path)
+            else:
+                continue
+
+        vid_idx = len(video_paths)
+        if row_heights[row_idx] == "":
+            row_heights[row_idx] = "h{}".format(vid_idx)
+        input_filter_part = "[{}:v]scale={}:-1,{}[v{}]".format(vid_idx, w, pad_str, vid_idx)
+        input_filter_parts.append(input_filter_part)
+        xstack_input_part = "[v{}]".format(vid_idx)
+        xstack_input_parts.append(xstack_input_part)
         video_paths.append(video_path)
+        layout_w = "+".join(["w{}".format(i) for i in range(col_idx)]) if col_idx > 0 else "0"
+        layout_h = "+".join(row_heights[:row_idx]) if row_idx > 0 else "0"
+        layout_part = layout_w + "_" + layout_h
+        layout_parts.append(layout_part)
 
-inputs_str = " ".join(["-i " + video_path for video_path in video_paths if video_path])
-num_inputs = len(inputs_str.split("-i")) - 1
-
-crop_str = "crop=trunc(iw/2)*2:trunc(ih/2)*2"
-pad_str = "pad=ceil(iw/2)*2+4:ceil(ih/2)*2+4:2:2"
-
-# empty_elem_str = "drawbox=t=fill:c=black,scale={}:{}[v{}]"
-
-w = selected_dataset["width"]
-input_filter_parts = []
-xstack_input_parts = []
-input_vid_no = 0
-for i in range(num_elements):
-    row_idx = i // num_cols
-    xstack_input_part = "[v{}]".format(i)
-    xstack_input_parts.append(xstack_input_part)
-    if video_paths[i]:
-        input_filter_part = "[{}:v]".format(input_vid_no)
-        input_vid_no += 1
-    else:
-        input_filter_part = "color=c=black,"
-        input_filter_part = "[{}]drawbox=:w={}:h={}:color=black:t=fill,".format(num_inputs, w, row_heights[row_idx])
-        pass
-    input_filter_part = input_filter_part + "scale={}:-1,{}[v{}]".format(w, pad_str, i)
-    input_filter_parts.append(input_filter_part)
+inputs_str = " ".join(["-i " + video_path for video_path in video_paths])
+num_inputs = len(video_paths)
 
 input_scaling_str = ";".join(input_filter_parts)
 xstack_input_str = "".join(xstack_input_parts)
-
-layout = Layout(num_rows, num_cols)
-layout_str = layout.get_layout_str()
+layout_str = "|".join(layout_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 + ":fill=black\"" + opt + " output.mp4"
+opt_fill = ":fill=black"
+# opt_fill = ""
+ffmpeg_command_str = "ffmpeg -y " + inputs_str + " -filter_complex \"" + input_scaling_str + ";" + xstack_input_str + "xstack=inputs=" + str(num_inputs) + ":layout=" + layout_str + opt_fill + "\"" + opt + " output.mp4"
 print(ffmpeg_command_str)
 ret = subprocess.call(ffmpeg_command_str, shell=True)
 

utils.py

@@ -41,39 +41,6 @@ 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
-
-
 def get_video_height(video_path):
     probe = ffmpeg.probe(video_path)
     video_streams = ffmpeg.probe(video_path, select_streams="v")