commit wip

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
+from utils import get_configs, get_display_names, get_path_for_viz, Layout, get_video_height
 
 # st.header("EVREAL  - Event-based Video Reconstruction Evaluation and Analysis Library")
 #
@@ -78,7 +78,7 @@ 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)
@@ -99,45 +99,82 @@ gt_viz.extend([viz for viz in gt_only_viz if viz['gt_type'] == 'event'])
 num_gt_rows = len(gt_viz)
 num_rows = max(num_model_rows, num_gt_rows)
 
-num_model_columns = len(selected_models)
+if len(gt_viz) > 0:
+    selected_models.append(ground_truth)
 
-num_elements = num_rows * num_model_columns
+num_cols = len(selected_models)
 
-layout = Layout(num_rows, num_model_columns)
-layout_str = layout.get_layout_str()
+num_elements = num_rows * num_cols
 
 video_paths = []
+row_heights = [0]*num_rows
 for row_idx in range(num_rows):
-    for col_idx in range(num_model_columns):
-        video_path = get_path_for_viz(data_base_path, selected_dataset, selected_sequence,
-                                      selected_models[col_idx], model_viz[row_idx])
+    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:
+                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:
+                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)
         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])
+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_scaling_parts = []
+input_filter_parts = []
 xstack_input_parts = []
+input_vid_no = 0
 for i in range(num_elements):
-    input_scaling_part = "[{}:v]scale={}:-1,{}[v{}]".format(i, w, pad_str, i)
-    input_scaling_parts.append(input_scaling_part)
+    row_idx = i // num_cols
     xstack_input_part = "[v{}]".format(i)
     xstack_input_parts.append(xstack_input_part)
-input_scaling_str = ";".join(input_scaling_parts)
+    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)
+
+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()
+
 # 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"
+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"
 print(ffmpeg_command_str)
-subprocess.call(ffmpeg_command_str, shell=True)
+ret = subprocess.call(ffmpeg_command_str, shell=True)
+
+if ret != 0:
+    st.error("Error while generating video.")
+    st.stop()
+
+video_file = open('output.mp4', 'rb')
+video_bytes = video_file.read()
+
+st.video(video_bytes)

utils.py

@@ -3,6 +3,7 @@ import glob
 import json
 from pathlib import Path
 from collections import OrderedDict
+import ffmpeg
 
 
 def read_json(fname):
@@ -71,3 +72,9 @@ class Layout:
                 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")
+    return video_streams['streams'][0]['height']