hw4: done task 3_2

hw4/README.md

+
+Задание 1 [Ссылка на скринкаст](https://drive.google.com/file/d/1NkOrWvB7MN_F7HE4JrXJYbcaADkSU-BT/view?usp=sharing)
+
+Задание 2 [Ссылка на скринкаст](https://drive.google.com/file/d/1FZa6tnDL8EPwTQNkvOc7InXlRpPxhXbs/view?usp=sharing)
+
+Задание 3 [Ссылка на архив с фото](https://drive.google.com/file/d/1lzfjR7TEkcoXOj29QZ4UbrwVrywpHwCs/view?usp=sharing)

hw4/requirements.txt

-numpy==2.2.2
-opencv-python==4.11.0.86
\ No newline at end of file
+absl-py==2.1.0
+altair==5.5.0
+astunparse==1.6.3
+attrs==25.1.0
+beautifulsoup4==4.12.3
+blinker==1.9.0
+cachetools==5.5.1
+certifi==2025.1.31
+charset-normalizer==3.4.1
+click==8.1.8
+contourpy==1.3.1
+cycler==0.12.1
+deepface==0.0.93
+faiss-cpu==1.10.0
+filelock==3.17.0
+fire==0.7.0
+Flask==3.1.0
+Flask-Cors==5.0.0
+flatbuffers==25.1.24
+fonttools==4.55.8
+fsspec==2024.12.0
+gast==0.6.0
+gdown==5.2.0
+gitdb==4.0.12
+GitPython==3.1.44
+google-pasta==0.2.0
+grpcio==1.70.0
+gunicorn==23.0.0
+h5py==3.12.1
+idna==3.10
+itsdangerous==2.2.0
+Jinja2==3.1.5
+joblib==1.4.2
+jsonschema==4.23.0
+jsonschema-specifications==2024.10.1
+keras==3.8.0
+kiwisolver==1.4.8
+libclang==18.1.1
+lz4==4.4.3
+Markdown==3.7
+markdown-it-py==3.0.0
+MarkupSafe==3.0.2
+matplotlib==3.10.0
+mdurl==0.1.2
+ml-dtypes==0.4.1
+mpmath==1.3.0
+mtcnn==1.0.0
+namex==0.0.8
+narwhals==1.24.1
+networkx==3.4.2
+numpy==1.26.4
+opencv-python==4.11.0.86
+opt_einsum==3.4.0
+optree==0.14.0
+packaging==24.2
+pandas==2.2.3
+pillow==11.1.0
+protobuf==4.25.6
+psutil==6.1.1
+py-cpuinfo==9.0.0
+pyarrow==19.0.0
+pydeck==0.9.1
+Pygments==2.19.1
+pyparsing==3.2.1
+PySocks==1.7.1
+python-dateutil==2.9.0.post0
+pytz==2025.1
+PyYAML==6.0.2
+referencing==0.36.2
+requests==2.32.3
+retina-face==0.0.17
+rich==13.9.4
+rpds-py==0.22.3
+scipy==1.15.1
+seaborn==0.13.2
+six==1.17.0
+smmap==5.0.2
+soupsieve==2.6
+streamlit==1.41.1
+sympy==1.13.1
+tenacity==9.0.0
+tensorboard==2.18.0
+tensorboard-data-server==0.7.2
+tensorflow==2.18.0
+tensorflow-io-gcs-filesystem==0.37.1
+tensorflow-metal==1.2.0
+termcolor==2.5.0
+tf_keras==2.18.0
+toml==0.10.2
+torch==2.6.0
+torchvision==0.21.0
+tornado==6.4.2
+tqdm==4.67.1
+typing_extensions==4.12.2
+tzdata==2025.1
+Unidecode==1.3.8
+urllib3==2.3.0
+Werkzeug==3.1.3
+wrapt==1.17.2

hw4/task1.py

@@ -22,7 +22,7 @@ def visualize(frame, faces, fps, recognizer, ground_base_feature):
             
             cosine_similarity_threshold = 0.363
             l2_similarity_threshold = 1.128
-         
+
             cosine_score = recognizer.match(
                 face_feature,
                 ground_base_feature,
@@ -76,7 +76,6 @@ def get_ground_base_feature(detector, recognizer):
     return face_feature
 
 
-
 def main():
     detector = cv.FaceDetectorYN.create(
         env["face_detection_model"],

hw4/task2.py

@@ -30,7 +30,7 @@ def create_dataset_embeddings() -> (np.ndarray, List):
     for path in Path("task2_input/hse_faces_miem/").rglob("*.jpeg"):
         path = str(path)
         print(path)
-        image = np.array(Image.open(path))[:, :, ::-1]
+        image = np.array(Image.open(path))
 
         try:
             detections = get_image_embeddings(image)
@@ -102,7 +102,7 @@ def main(index, ids):
 
             np_image = np.asarray(
                 Image.open(uploaded_file)
-            )[:, :, ::-1]
+            )
 
             detection = get_image_embeddings(np_image)[0]
 
@@ -127,7 +127,8 @@ def main(index, ids):
                 detected_name,
             )
 
-    st.markdown("""
+    st.markdown(
+        """
         <style>
             .stButton>button {
                 width: 100%;

hw4/task3_1.py

@@ -69,7 +69,7 @@ def main():
             if image is None:
                 break
             result_image = image.copy()
-            np_image = np.asarray(image[:, :])[:, :, ::-1]
+            np_image = np.asarray(image[:, :])
             print("Processing frame: ", current_frame_index)
             frame_filename = os.path.join(video_output_dir, f"frame_{current_frame_index}.jpg")
 
@@ -99,7 +99,7 @@ def main():
                     y2 = face["facial_area"]["h"] + y
 
                     try:
-                        verified |= DeepFace.verify(person_path, np_image[y:y2, x:x2])["verified"]
+                        verified = DeepFace.verify(person_path, np_image[y:y2, x:x2])["verified"]
                     except Exception as e:
                         print(f"Failed to verify person {person_path} with exception {e}")
                     if verified:

hw4/task3_1_output/results.txt

@@ -2,7 +2,7 @@
 Video: palmasizade_video.mp4
 Labels: [[0, 12180, ['palmasizade']], [12210, 12270, ['palmasizade', 'vkrepker']], [12300, 18000, ['palmasizade']]]
 Length: 17751
-True Positives: 53
-False Positives: 303
-Precision: 0.14887640449438203
-Correct Frames: [2700, 3000, 7600, 7650, 7700, 7750, 8400, 8500, 8950, 9100, 9200, 9250, 9300, 9350, 9400, 9450, 9500, 9550, 9600, 9650, 9700, 9750, 10200, 10400, 10450, 10700, 10950, 11000, 11200, 11250, 11350, 11400, 11450, 11500, 11550, 11600, 11900, 12250, 12300, 12350, 12400, 12450, 12600, 12650, 12800, 12850, 12900, 13700, 13900, 14350, 14800, 15500, 15550]
+True Positives: 54
+False Positives: 302
+Precision: 0.15168539325842698
+Correct Frames: [2700, 3000, 7600, 7650, 7700, 7750, 7800, 8000, 9100, 9150, 9200, 9250, 9300, 9400, 9450, 9500, 9550, 9600, 9650, 9700, 9750, 9800, 10400, 10450, 10700, 10950, 11000, 11200, 11250, 11350, 11400, 11450, 11500, 11550, 11600, 11700, 11900, 12250, 12300, 12350, 12400, 12450, 12600, 12800, 12850, 12900, 13700, 13800, 13900, 14350, 14800, 15500, 15550, 16650]

hw4/task3_2.py

+from deepface import DeepFace
+import pandas as pd
+import numpy as np
+import cv2 as cv
+import os
+
+models = [
+    "VGG-Face",
+    "Facenet",
+    "Facenet512",
+    "OpenFace",
+    # "DeepFace",
+    "DeepID",
+    "ArcFace",
+    # "Dlib",
+    "SFace",
+    "GhostFaceNet"
+]
+
+results = {
+    "model": models,
+    "original": [],
+    "rotated_45": [],
+    "rotated_90": [],
+    "blurred": [],
+    "dark": [],
+    "bright": [],
+    "noisy": []
+}
+
+
+def augment_dataset():
+    dataset_folder = "./task3_2_input/dataset"
+    original_folder = os.path.join(dataset_folder, "original")
+    for filename in os.listdir(original_folder):
+        image_path = os.path.join(original_folder, filename)
+        image = cv.imread(image_path)
+        if image is not None:
+            print(f"Processing {filename}")
+
+            # Rotate 45 degrees
+            (h, w) = image.shape[:2]
+            center = (w // 2, h // 2)
+            M = cv.getRotationMatrix2D(center, 45, 1.0)
+            rotated_45 = cv.warpAffine(image, M, (w, h))
+
+            os.makedirs(os.path.join(dataset_folder, "rotated_45"), exist_ok=True)
+            rotated_45_path = os.path.join(
+                dataset_folder,
+                "rotated_45",
+                f"{os.path.splitext(filename)[0]}_rotated_45{os.path.splitext(filename)[1]}"
+            )
+            cv.imwrite(rotated_45_path, rotated_45)
+
+            # Rotate 90 degrees
+            rotated_90 = cv.rotate(image, cv.ROTATE_90_CLOCKWISE)
+            os.makedirs(os.path.join(dataset_folder, "rotated_90"), exist_ok=True)
+            rotated_90_path = os.path.join(
+                dataset_folder,
+                "rotated_90",
+                f"{os.path.splitext(filename)[0]}_rotated_90{os.path.splitext(filename)[1]}"
+            )
+            cv.imwrite(rotated_90_path, rotated_90)
+
+            # Blur
+            blurred = cv.GaussianBlur(image, (15, 15), 0)
+            os.makedirs(os.path.join(dataset_folder, "blurred"), exist_ok=True)
+            blurred_path = os.path.join(
+                dataset_folder,
+                "blurred",
+                f"{os.path.splitext(filename)[0]}_blurred{os.path.splitext(filename)[1]}"
+            )
+            cv.imwrite(blurred_path, blurred)
+
+            # Brightness
+            # Decrease brightness by 50%
+            dark = cv.convertScaleAbs(image, alpha=0.5, beta=0)
+            os.makedirs(os.path.join(dataset_folder, "dark"), exist_ok=True)
+            dark_path = os.path.join(
+                dataset_folder,
+                "dark",
+                f"{os.path.splitext(filename)[0]}_dark{os.path.splitext(filename)[1]}"
+            )
+            cv.imwrite(dark_path, dark)
+
+            # Increase brightness by 50%
+            bright = cv.convertScaleAbs(image, alpha=1.5, beta=0)
+            os.makedirs(os.path.join(dataset_folder, "bright"), exist_ok=True)
+            bright_path = os.path.join(
+                dataset_folder,
+                "bright",
+                f"{os.path.splitext(filename)[0]}_bright{os.path.splitext(filename)[1]}"
+            )
+            cv.imwrite(bright_path, bright)
+
+            # Add noise
+            noise = image.copy()
+            mean = 0
+            sigma = 25
+            gauss = np.random.normal(mean, sigma, noise.shape).astype("uint8")
+            noisy_image = cv.add(noise, gauss)
+            os.makedirs(os.path.join(dataset_folder, "noisy"), exist_ok=True)
+            noisy_path = os.path.join(
+                dataset_folder,
+                "noisy",
+                f"{os.path.splitext(filename)[0]}_noisy{os.path.splitext(filename)[1]}"
+            )
+            cv.imwrite(noisy_path, noisy_image)
+        else:
+            print(f"Failed to load {filename}")
+
+
+def main():
+    # augment_dataset()
+    
+    for model in models:
+        print("==================================")
+        print(f"Processing model {model}")
+
+        for dataset in list(results.keys())[1:]:
+            detection = DeepFace.find(
+                img_path="task3_2_input/egor.jpg",
+                db_path=f"task3_2_input/dataset/{dataset}",
+                model_name=model,
+                enforce_detection=False
+            )
+            recall = float(len(detection[0])) / 20.0
+            results[dataset].append(recall)
+
+            print(len(detection[0]))
+
+        print()
+
+    table = pd.DataFrame(results)
+    print(table)
+    table.to_csv("task3_2_output/results.txt", index=False)
+
+
+if __name__ == "__main__":
+    main()

hw4/task3_2_output/results.txt

+model,original,rotated_45,rotated_90,blurred,dark,bright,noisy
+VGG-Face,0.4,0.05,0.05,0.25,0.5,0.4,0.0
+Facenet,0.1,0.0,0.05,0.05,0.0,0.1,0.0
+Facenet512,0.0,0.0,0.05,0.0,0.0,0.0,0.0
+OpenFace,0.0,0.0,0.0,0.0,0.1,0.0,0.0
+DeepID,0.1,0.05,0.0,0.05,0.0,0.1,0.05
+ArcFace,0.6,0.0,0.05,0.35,0.6,0.5,0.0
+SFace,0.4,0.0,0.05,0.2,0.45,0.35,0.0
+GhostFaceNet,0.25,0.0,0.05,0.1,0.15,0.2,0.0