working

.gitignore

@@ -5,3 +5,6 @@
 # data
 assets
 out
+
+# python
+__pycache__

README.md

@@ -1,13 +1,13 @@
 ---
 title: Aero Recognize
-emoji: 📊
+emoji: 🛫
 colorFrom: gray
-colorTo: green
+colorTo: indigo
 sdk: gradio
 sdk_version: 4.12.0
-app_file: app.py
+app_file: app/main.py
 pinned: false
 license: bsd-2-clause
 ---
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Aero Recognize
+Aeroplane detector and action classifier.

playgrounds/load_video.py

@@ -4,7 +4,6 @@ import numpy as np
 import tensorflow as tf
 import cv2
 from pathlib import Path
-print('Modules loaded.')
 
 SPLIT_RATIO = 0.7
 BATCH_SIZE = 8
@@ -64,13 +63,15 @@ def frames_from_video_file(video_path, n_frames, output_size=(224, 224), frame_s
 
     src.set(cv2.CAP_PROP_POS_FRAMES, start)
     # ret is a boolean indicating whether read was successful, frame is the image itself
-    ret, frame = src.read()
+    ok, frame = src.read()
+    if not ok:
+        raise ValueError('read video not success')
     result.append(format_frames(frame, output_size))
 
     for _ in range(n_frames - 1):
         for _ in range(frame_step):
-            ret, frame = src.read()
-        if ret:
+            ok, frame = src.read()
+        if ok:
             frame = format_frames(frame, output_size)
             result.append(frame)
         else:
@@ -109,5 +110,3 @@ def frame_generator(data_dir: Path, n_frames: int, split: Literal['training', 'v
             label = class_ids_for_name[name]  # Encode labels
             yield video_frames, label
     return generator
-
-main()

playgrounds/main.py

@@ -0,0 +1,3 @@
+from playgrounds.yolo import main
+
+main()

playgrounds/movinet.py

@@ -2,7 +2,6 @@ import tensorflow as tf
 import numpy as np
 import tensorflow_hub as hub
 import keras
-print('Modules loaded.')
 
 labels_path = keras.utils.get_file(
     fname='labels.txt',
@@ -14,7 +13,6 @@ with open(labels_path, 'r', encoding='utf-8') as file:
 
 KINETICS_600_LABELS = np.array([line.strip() for line in lines])
 KINETICS_600_LABELS[:20]
-print('Labels loaded.')
 
 def main():
     jumping_jack_path = 'assets/jumping_pack.gif'
@@ -80,5 +78,3 @@ def load_gif(file_path, image_size=(224, 224)):
     # ref: https://www.tensorflow.org/hub/common_signatures/images#input
     video = tf.cast(video, tf.float32) / 255.
     return video
-
-main()

playgrounds/yolo.py

@@ -2,35 +2,39 @@ import keras
 import keras_cv
 import numpy as np
 import tensorflow as tf
-print('Modules loaded.')
 
-pretrained_model = keras_cv.models.YOLOV8Detector.from_preset(
-    "yolo_v8_m_pascalvoc", bounding_box_format="xywh"
-)
-print('Model loaded.')
+from playgrounds.load_video import frames_from_video_file
 
-inference_resizing = keras_cv.layers.Resizing(
-    640, 640, pad_to_aspect_ratio=True, bounding_box_format="xywh"
-)
+def main():
+    pretrained_model = keras_cv.models.YOLOV8Detector.from_preset(
+        "yolo_v8_m_pascalvoc", bounding_box_format="xywh"
+    )
+    print('Model loaded.')
 
-class_ids = [
-    "Aeroplane", "Bicycle", "Bird", "Boat", "Bottle", "Bus", "Car", "Cat", "Chair", "Cow", "Dining Table",
-    "Dog", "Horse", "Motorbike", "Person", "Potted Plant", "Sheep", "Sofa", "Train", "Tvmonitor", "Total",
-]
-class_mapping = {i: c for (i, c) in enumerate(class_ids)}
+    inference_resizing = keras_cv.layers.Resizing(
+        640, 640, pad_to_aspect_ratio=True, bounding_box_format="xywh"
+    )
 
-raw = tf.io.read_file('assets/IMG_9528.gif')
-video = tf.io.decode_gif(raw)
-image = video[0]
-file = tf.io.encode_png(image)
-tf.io.write_file('out/t.png', file)
-# image = keras.utils.load_img('assets/nick-morales-BwYcH78rcpI-unsplash.jpg')
-# image = np.array(image)
+    class_ids = [
+        "Aeroplane", "Bicycle", "Bird", "Boat", "Bottle", "Bus", "Car", "Cat", "Chair", "Cow", "Dining Table",
+        "Dog", "Horse", "Motorbike", "Person", "Potted Plant", "Sheep", "Sofa", "Train", "Tvmonitor", "Total",
+    ]
+    class_mapping = {i: c for (i, c) in enumerate(class_ids)}
 
-image_batch = inference_resizing([image])
+    # raw = tf.io.read_file('assets/IMG_9528.gif')
+    # video = tf.io.decode_gif(raw)
+    video = frames_from_video_file('assets/dataset/Flying/2kNjmM8BnD0_230.0_238.0.mp4', 3, (640,640))
+    image = video[0]
+    image = (image*255).astype(np.uint8)
+    file = tf.io.encode_png(image)
+    tf.io.write_file('out/t.png', file)
+    # image = keras.utils.load_img('assets/nick-morales-BwYcH78rcpI-unsplash.jpg')
+    # image = np.array(image)
 
-y_pred = pretrained_model.predict(image_batch)
-classes = y_pred['classes']
-boxes = y_pred["boxes"]
-print(f'Classes: {classes}')
-print(f'Boxes: {boxes}')
+    image_batch = inference_resizing([image])
+
+    y_pred = pretrained_model.predict(image_batch)
+    classes = y_pred['classes']
+    boxes = y_pred["boxes"]
+    print(f'Classes: {classes}')
+    print(f'Boxes: {boxes}')