show_examples_from_index now also works on index_ser.txt files

training/show_examples_from_index.py

@@ -1,21 +1,54 @@
 # ruff: noqa: T201
 
+import argparse
 import os
-import sys
 from typing import Any
 
 import cv2
 import numpy as np
 from termcolor import colored
 
-index_file_name = sys.argv[1]
-number_of_samples_per_iteration = int(sys.argv[2])
+parser = argparse.ArgumentParser(description="Show examples from a dataset index")
+parser.add_argument("index", type=str, help="Index file name")
+parser.add_argument("number_of_images", type=int, help="Number of images to show at once")
+parser.add_argument(
+    "--sorted",
+    action="store_true",
+    help="Show the images in the order they are in the index file",
+)
+parser.add_argument(
+    "--min-ser",
+    type=int,
+    help="Minimum SER to show",
+)
+parser.add_argument(
+    "--max-ser",
+    type=int,
+    help="Maximum SER to show",
+)
+args = parser.parse_args()
+
+index_file_name = args.index
+number_of_samples_per_iteration = int(args.number_of_images)
 
 index_file = open(index_file_name)
 index_lines = index_file.readlines()
 index_file.close()
 
-np.random.shuffle(index_lines)
+ser_position = 2
+
+if not args.sorted:
+    np.random.shuffle(index_lines)
+
+if args.min_ser is not None:
+    index_lines = [
+        line for line in index_lines if int(line.split(",")[ser_position]) >= args.min_ser
+    ]
+
+if args.max_ser is not None:
+    index_lines = [
+        line for line in index_lines if int(line.split(",")[ser_position]) <= args.max_ser
+    ]
 
 
 def print_color(text: str, highlights: list[str], color: Any) -> None:
@@ -29,10 +62,11 @@ def print_color(text: str, highlights: list[str], color: Any) -> None:
 
 
 while True:
-    batch = []
-    for _ in range(number_of_samples_per_iteration):
+    batch: list[str] = []
+    while len(batch) < number_of_samples_per_iteration:
         if len(index_lines) == 0:
             break
+
         batch.append(index_lines.pop())
 
     if len(batch) == 0:
@@ -45,7 +79,12 @@ def print_color(text: str, highlights: list[str], color: Any) -> None:
     print("==========================================")
     print()
     for line in batch:
-        image_path, semantic_path = line.strip().split(",")
+        cells = line.strip().split(",")
+        image_path = cells[0]
+        semantic_path = cells[1]
+        ser: None | int = None
+        if len(cells) > ser_position:
+            ser = int(cells[ser_position])
         agnostic_path = semantic_path.replace(".semantic", ".agnostic")
         image = cv2.imread(image_path)
         with open(semantic_path) as file:
@@ -60,7 +99,10 @@ def print_color(text: str, highlights: list[str], color: Any) -> None:
         else:
             images = np.concatenate((images, image), axis=0)
         print()
-        print(">>> " + image_path)
+        if ser is not None:
+            print(">>> " + image_path + f" SER: {ser}%")
+        else:
+            print(">>> " + image_path)
         print_color(semantic, ["barline", "#", "N", "b"], "green")
     cv2.imshow("Images", images)  # type: ignore
     escKey = 27

validation/ser_for_data_set.py

@@ -17,9 +17,6 @@ def calc_symbol_error_rate_for_list(dataset: list[str], result_file: str, config
     total = len(dataset)
 
     with open(result_file, "w") as result:
-        result.write(
-            "img_path,ser,len_expected,len_actual,added_symbols,missing_symbols,expected,actual\n"
-        )
         for sample in dataset:
             img_path, semantic_path = sample.strip().split(",")
             expected_str = _load_semantic_file(semantic_path)[0].strip()