Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -6,24 +6,113 @@ Spaces for showing the model usage.
|
|
6 |
Author:
|
7 |
- Thomas Chaigneau @ChainYo
|
8 |
"""
|
|
|
|
|
9 |
import gradio as gr
|
|
|
|
|
|
|
|
|
10 |
|
11 |
from huggingface_hub import from_pretrained_keras
|
12 |
|
13 |
|
14 |
-
|
15 |
-
|
16 |
-
Inference function.
|
17 |
-
"""
|
18 |
|
19 |
model = from_pretrained_keras("ChainYo/video-classification-cnn-rnn")
|
20 |
-
samples =
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
|
|
22 |
app = gr.Interface(
|
23 |
-
|
24 |
-
inputs=[],
|
25 |
outputs=[],
|
26 |
title="Keras Video Classification CNN-RNN model",
|
27 |
description="Keras Working Group",
|
|
|
28 |
examples=samples
|
29 |
).launch(enable_queue=True, cache_examples=True)
|
|
|
6 |
Author:
|
7 |
- Thomas Chaigneau @ChainYo
|
8 |
"""
|
9 |
+
import os
|
10 |
+
import cv2
|
11 |
import gradio as gr
|
12 |
+
import numpy as np
|
13 |
+
|
14 |
+
from tensorflow import keras
|
15 |
+
from tensorflow_docs.vis import embed
|
16 |
|
17 |
from huggingface_hub import from_pretrained_keras
|
18 |
|
19 |
|
20 |
+
IMG_SIZE = 224
|
21 |
+
NUM_FEATURES = 2048
|
|
|
|
|
22 |
|
23 |
model = from_pretrained_keras("ChainYo/video-classification-cnn-rnn")
|
24 |
+
samples = []
|
25 |
+
for file in os.listdir("samples"):
|
26 |
+
print(file)
|
27 |
+
tag = file.split("_")[1]
|
28 |
+
samples.append([f"samples/{file}", 25])
|
29 |
+
|
30 |
+
|
31 |
+
def crop_center_square(frame):
|
32 |
+
y, x = frame.shape[0:2]
|
33 |
+
min_dim = min(y, x)
|
34 |
+
start_x = (x // 2) - (min_dim // 2)
|
35 |
+
start_y = (y // 2) - (min_dim // 2)
|
36 |
+
return frame[start_y : start_y + min_dim, start_x : start_x + min_dim]
|
37 |
+
|
38 |
+
|
39 |
+
def load_video(path, max_frames=0, resize=(IMG_SIZE, IMG_SIZE)):
|
40 |
+
cap = cv2.VideoCapture(path)
|
41 |
+
frames = []
|
42 |
+
try:
|
43 |
+
while True:
|
44 |
+
ret, frame = cap.read()
|
45 |
+
if not ret:
|
46 |
+
break
|
47 |
+
frame = crop_center_square(frame)
|
48 |
+
frame = cv2.resize(frame, resize)
|
49 |
+
frame = frame[:, :, [2, 1, 0]]
|
50 |
+
frames.append(frame)
|
51 |
+
|
52 |
+
if len(frames) == max_frames:
|
53 |
+
break
|
54 |
+
finally:
|
55 |
+
cap.release()
|
56 |
+
return np.array(frames)
|
57 |
+
|
58 |
+
|
59 |
+
def build_feature_extractor():
|
60 |
+
feature_extractor = keras.applications.InceptionV3(
|
61 |
+
weights="imagenet",
|
62 |
+
include_top=False,
|
63 |
+
pooling="avg",
|
64 |
+
input_shape=(IMG_SIZE, IMG_SIZE, 3),
|
65 |
+
)
|
66 |
+
preprocess_input = keras.applications.inception_v3.preprocess_input
|
67 |
+
|
68 |
+
inputs = keras.Input((IMG_SIZE, IMG_SIZE, 3))
|
69 |
+
preprocessed = preprocess_input(inputs)
|
70 |
+
|
71 |
+
outputs = feature_extractor(preprocessed)
|
72 |
+
return keras.Model(inputs, outputs, name="feature_extractor")
|
73 |
+
|
74 |
+
|
75 |
+
feature_extractor = build_feature_extractor()
|
76 |
+
|
77 |
+
def prepare_video(frames, max_seq_length: int = 20):
|
78 |
+
frames = frames[None, ...]
|
79 |
+
frame_mask = np.zeros(shape=(1, max_seq_length,), dtype="bool")
|
80 |
+
frame_features = np.zeros(shape=(1, max_seq_length, NUM_FEATURES), dtype="float32")
|
81 |
+
|
82 |
+
for i, batch in enumerate(frames):
|
83 |
+
video_length = batch.shape[0]
|
84 |
+
length = min(max_seq_length, video_length)
|
85 |
+
for j in range(length):
|
86 |
+
frame_features[i, j, :] = feature_extractor.predict(batch[None, j, :])
|
87 |
+
frame_mask[i, :length] = 1 # 1 = not masked, 0 = masked
|
88 |
+
|
89 |
+
return frame_features, frame_mask
|
90 |
+
|
91 |
+
|
92 |
+
def sequence_prediction(path):
|
93 |
+
class_vocab = ["CricketShot", "PlayingCello", "Punch", "ShavingBeard", "TennisSwing"]
|
94 |
+
|
95 |
+
frames = load_video(path)
|
96 |
+
frame_features, frame_mask = prepare_video(frames)
|
97 |
+
probabilities = model.predict([frame_features, frame_mask])[0]
|
98 |
+
|
99 |
+
for i in np.argsort(probabilities)[::-1]:
|
100 |
+
print(f" {class_vocab[i]}: {probabilities[i] * 100:5.2f}%")
|
101 |
+
return frames
|
102 |
+
|
103 |
+
|
104 |
+
def to_gif(images):
|
105 |
+
converted_images = images.astype(np.uint8)
|
106 |
+
return embed.embed_file(converted_images, format="gif")
|
107 |
+
|
108 |
|
109 |
+
article = article = "<div style='text-align: center;'><a href='https://github.com/ChainYo' target='_blank'>Space by Thomas Chaigneau</a><br><a href='https://keras.io/examples/vision/video_classification/' target='_blank'>Keras example by Sayak Paul</a></div>"
|
110 |
app = gr.Interface(
|
111 |
+
sequence_prediction,
|
112 |
+
inputs=[gr.inputs.Video(label="Video", type="mp4")],
|
113 |
outputs=[],
|
114 |
title="Keras Video Classification CNN-RNN model",
|
115 |
description="Keras Working Group",
|
116 |
+
article=article,
|
117 |
examples=samples
|
118 |
).launch(enable_queue=True, cache_examples=True)
|