Test script to check overall accuracy compared to test cases generated by WSJT-X

utils/run_tests.py

@@ -0,0 +1,49 @@
+#!/usr/bin/env python3
+
+import sys, os, subprocess
+
+def parse(line):
+    fields = line.split()
+    freq = fields[3]
+    dest = fields[5] if len(fields) > 5 else ''
+    source = fields[6] if len(fields) > 6 else ''
+    report = fields[7] if len(fields) > 7 else ''
+    return ' '.join([dest, source, report])
+
+wav_dir = sys.argv[1]
+wav_files = [os.path.join(wav_dir, f) for f in os.listdir(wav_dir)]
+wav_files = [f for f in wav_files if os.path.isfile(f) and os.path.splitext(f)[1] == '.wav']
+txt_files = [os.path.splitext(f)[0] + '.txt' for f in wav_files]
+
+n_extra = 0
+n_missed = 0
+n_total = 0
+for wav_file, txt_file in zip(wav_files, txt_files):
+    result = subprocess.run(['./decode_ft8', wav_file], stdout=subprocess.PIPE)
+    result = result.stdout.decode('utf-8').split('\n')
+    result = [parse(x) for x in result if len(x) > 0]
+    #print(result[0])
+    result = set(result)
+    
+    expected = open(txt_file).read().split('\n')
+    expected = [parse(x) for x in expected if len(x) > 0]
+    #print(expected[0])
+    expected = set(expected)
+    
+    extra_decodes = result - expected
+    missed_decodes = expected - result
+    print(len(result), '/', len(expected))
+    if len(extra_decodes) > 0:
+        print('Extra decodes: ', list(extra_decodes))
+    #print('Missed decodes: ', list(missed_decodes))
+    
+    n_total += len(expected)
+    n_extra += len(extra_decodes)
+    n_missed += len(missed_decodes)
+    
+    #break
+
+print('Total: %d, extra: %d (%.1f%%), missed: %d (%.1f%%)' % 
+        (n_total, n_extra, 100.0*n_extra/n_total, n_missed, 100.0*n_missed/n_total))
+recall = (n_total - n_missed) / float(n_total)
+print('Recall: %.1f%%' % (100*recall, ))