add tests (#18)

* add tests

* add sentencepiece dependency

manga_ocr_dev/requirements.txt

@@ -11,6 +11,7 @@ numpy
 opencv-python
 pandas
 Pillow
+pytest
 scikit-image
 scikit-learn
 scipy

setup.py

@@ -26,6 +26,7 @@ setup(
         "numpy",
         "Pillow",
         "pyperclip",
+        "sentencepiece",
         "torch>=1.0",
         "transformers>=4.12.5",
         "unidic_lite",

tests/data/expected_results.json

@@ -0,0 +1,50 @@
+[
+  {
+    "filename": "00.jpg",
+    "result": "素直にあやまるしか"
+  },
+  {
+    "filename": "01.jpg",
+    "result": "立川で見た〝穴〟の下の巨大な眼は："
+  },
+  {
+    "filename": "02.jpg",
+    "result": "実戦剣術も一流です"
+  },
+  {
+    "filename": "03.jpg",
+    "result": "第３０話重苦しい闇の奥で静かに呼吸づきながら"
+  },
+  {
+    "filename": "04.jpg",
+    "result": "きのうハンパーヶとって、ゴメン！！！"
+  },
+  {
+    "filename": "05.jpg",
+    "result": "ぎゃっ"
+  },
+  {
+    "filename": "06.jpg",
+    "result": "ピンポーーン"
+  },
+  {
+    "filename": "07.jpg",
+    "result": "ＬＩＮＫ！私達７人の力でガノンの塔の結界をやぶります"
+  },
+  {
+    "filename": "08.jpg",
+    "result": "ファイアパンチ"
+  },
+  {
+    "filename": "09.jpg",
+    "result": "少し黙っている"
+  },
+  {
+    "filename": "10.jpg",
+    "result": "わかるかな〜？"
+  },
+  {
+    "filename": "11.jpg",
+    "result": "警察にも先生にも町中の人達に！！"
+  }
+]

tests/generate_expected_results.py

@@ -0,0 +1,25 @@
+import json
+from pathlib import Path
+
+from tqdm import tqdm
+
+from manga_ocr import MangaOcr
+
+TEST_DATA_ROOT = Path(__file__).parent / 'data'
+
+
+def generate_expected_results():
+    mocr = MangaOcr()
+
+    results = []
+
+    for path in tqdm(sorted((TEST_DATA_ROOT / 'images').iterdir())):
+        result = mocr(path)
+        results.append({'filename': path.name, 'result': result})
+
+    (TEST_DATA_ROOT / 'expected_results.json').write_text(json.dumps(results, ensure_ascii=False, indent=2),
+                                                          encoding='utf-8')
+
+
+if __name__ == '__main__':
+    generate_expected_results()

tests/test_ocr.py

@@ -0,0 +1,16 @@
+import json
+from pathlib import Path
+
+from manga_ocr import MangaOcr
+
+TEST_DATA_ROOT = Path(__file__).parent / 'data'
+
+
+def test_ocr():
+    mocr = MangaOcr()
+
+    expected_results = json.loads((TEST_DATA_ROOT / 'expected_results.json').read_text(encoding='utf-8'))
+
+    for item in expected_results:
+        result = mocr(TEST_DATA_ROOT / 'images' / item['filename'])
+        assert result == item['result']