eval over all vps, result df

1 files changed, 23 insertions, 11 deletions

diff --git a/bjoern/videoanalyse/post_processing.py b/bjoern/videoanalyse/post_processing.py
index a8d37c4..6ffff1e 100644
--- a/bjoern/videoanalyse/post_processing.py
+++ b/bjoern/videoanalyse/post_processing.py
@@ -3,27 +3,39 @@
 import argparse
 from pathlib import Path
 from pprint import pprint
+import pandas as pd
 
 import utils
 
 argparser = argparse.ArgumentParser(description="OCR-Logfile evaluation")
-argparser.add_argument("vp_dir", help="VP Directory")
+argparser.add_argument("vp_dir", help="Directory with all VPs")
 
 args = argparser.parse_args()
 
 data_path = Path(args.vp_dir)
-video_path = next(data_path.glob("*.mkv"))
-ocr_path = data_path / "analysis_results.csv"
-log_path = data_path / f"{data_path.stem}.csv"
+all_vp = [x for x in data_path.iterdir() if x.is_dir()]
 
-df = utils.combine_ocr_logs(video_path, ocr_path, log_path)
-df = df.fillna('')
+vp_results = []
+for vp_path in all_vp:
+    video_path = next(vp_path.glob("*.mkv"))
+    ocr_path = vp_path / "analysis_results.csv"
+    log_path = vp_path / f"{vp_path.stem}.csv"
 
-df = utils.calc_levenshtein_distance(df)
+    df = utils.combine_ocr_logs(video_path, ocr_path, log_path)
+    df = df.fillna('')
+    df["vp_code"] = vp_path.stem
 
+    df = utils.calc_levenshtein_distance(df)
 
-url_groups = utils.group_urls(list(df["url"].values))
-pprint(len(url_groups))
+    url_groups = utils.group_urls(list(df["url"].values))
+    pprint(len(url_groups))
 
-df.to_csv(f"{data_path}/metrics.csv")
-utils.write_grouped_metrics(df, url_groups, data_path)
+    df.to_csv(f"{vp_path}/metrics.csv")
+    utils.write_grouped_metrics(df, url_groups, vp_path)
+
+    df = pd.read_csv(f"{vp_path}/metrics_grps.csv")
+
+    vp_results.append(df)
+
+evals = utils.evaluate_results(vp_results)
+pprint(evals)