Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -30,8 +30,8 @@ def load_image_into_numpy_array(path):
|
|
30 |
image = Image.open(BytesIO(image_data))
|
31 |
return pil_image_as_numpy_array(image)
|
32 |
|
33 |
-
def load_model():
|
34 |
-
download_dir = snapshot_download(
|
35 |
saved_model_dir = os.path.join(download_dir, "saved_model")
|
36 |
detection_model = tf.saved_model.load(saved_model_dir)
|
37 |
return detection_model
|
@@ -64,9 +64,40 @@ def predict2(image_np):
|
|
64 |
agnostic_mode=False,
|
65 |
line_thickness=2)
|
66 |
|
67 |
-
|
68 |
|
69 |
-
return
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
70 |
|
71 |
def detect_video(video):
|
72 |
# Create a video capture object
|
@@ -110,7 +141,10 @@ def detect_video(video):
|
|
110 |
|
111 |
label_id_offset = 0
|
112 |
REPO_ID = "apailang/mytfodmodel"
|
113 |
-
detection_model = load_model()
|
|
|
|
|
|
|
114 |
samples_folder = 'data'
|
115 |
# pil_image = Image.open(image_path)
|
116 |
# image_arr = pil_image_as_numpy_array(pil_image)
|
@@ -131,21 +165,32 @@ test10 = os.path.join(os.path.dirname(__file__), "data/test10.jpeg")
|
|
131 |
test11 = os.path.join(os.path.dirname(__file__), "data/test11.jpeg")
|
132 |
test12 = os.path.join(os.path.dirname(__file__), "data/test12.jpeg")
|
133 |
|
134 |
-
|
135 |
fn=predict,
|
136 |
inputs=gr.Image(type="pil"),
|
137 |
outputs=gr.Image(type="pil"),
|
138 |
-
title="
|
139 |
-
description="Upload a Image for prediction",
|
140 |
examples=[[test1],[test2],[test3],[test4],[test5],[test6],[test7],[test8],[test9],[test10],[test11],[test12],],
|
141 |
cache_examples=True
|
142 |
)#.launch(share=True)
|
143 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
144 |
a = os.path.join(os.path.dirname(__file__), "data/a.mp4") # Video
|
145 |
b = os.path.join(os.path.dirname(__file__), "data/b.mp4") # Video
|
146 |
c = os.path.join(os.path.dirname(__file__), "data/c.mp4") # Video
|
147 |
|
148 |
-
|
149 |
video_out_file = os.path.join(samples_folder,'detected' + '.mp4')
|
150 |
|
151 |
stt_demo = gr.Interface(
|
@@ -160,7 +205,7 @@ stt_demo = gr.Interface(
|
|
160 |
cache_examples=False
|
161 |
)
|
162 |
|
163 |
-
demo = gr.TabbedInterface([
|
164 |
|
165 |
if __name__ == "__main__":
|
166 |
demo.launch()
|
|
|
30 |
image = Image.open(BytesIO(image_data))
|
31 |
return pil_image_as_numpy_array(image)
|
32 |
|
33 |
+
def load_model(model_repo_id):
|
34 |
+
download_dir = snapshot_download(model_repo_id)
|
35 |
saved_model_dir = os.path.join(download_dir, "saved_model")
|
36 |
detection_model = tf.saved_model.load(saved_model_dir)
|
37 |
return detection_model
|
|
|
64 |
agnostic_mode=False,
|
65 |
line_thickness=2)
|
66 |
|
67 |
+
result_pil_img2 = tf.keras.utils.array_to_img(image_np_with_detections[0])
|
68 |
|
69 |
+
return result_pil_img2
|
70 |
+
|
71 |
+
def predict3(pilimg):
|
72 |
+
|
73 |
+
image_np = pil_image_as_numpy_array(pilimg)
|
74 |
+
return predict4(image_np)
|
75 |
+
|
76 |
+
def predict4(image_np):
|
77 |
+
|
78 |
+
results = detection_model2(image_np)
|
79 |
+
|
80 |
+
# different object detection models have additional results
|
81 |
+
result = {key:value.numpy() for key,value in results.items()}
|
82 |
+
|
83 |
+
label_id_offset = 0
|
84 |
+
image_np_with_detections = image_np.copy()
|
85 |
+
|
86 |
+
viz_utils.visualize_boxes_and_labels_on_image_array(
|
87 |
+
image_np_with_detections[0],
|
88 |
+
result['detection_boxes'][0],
|
89 |
+
(result['detection_classes'][0] + label_id_offset).astype(int),
|
90 |
+
result['detection_scores'][0],
|
91 |
+
category_index,
|
92 |
+
use_normalized_coordinates=True,
|
93 |
+
max_boxes_to_draw=200,
|
94 |
+
min_score_thresh=.60,
|
95 |
+
agnostic_mode=False,
|
96 |
+
line_thickness=2)
|
97 |
+
|
98 |
+
result_pil_img4 = tf.keras.utils.array_to_img(image_np_with_detections[0])
|
99 |
+
|
100 |
+
return result_pil_img4
|
101 |
|
102 |
def detect_video(video):
|
103 |
# Create a video capture object
|
|
|
141 |
|
142 |
label_id_offset = 0
|
143 |
REPO_ID = "apailang/mytfodmodel"
|
144 |
+
detection_model = load_model(REPO_ID)
|
145 |
+
REPO_ID2 = "apailang/mytfodmodeltuned"
|
146 |
+
detection_model2 = load_model(REPO_ID2)
|
147 |
+
|
148 |
samples_folder = 'data'
|
149 |
# pil_image = Image.open(image_path)
|
150 |
# image_arr = pil_image_as_numpy_array(pil_image)
|
|
|
165 |
test11 = os.path.join(os.path.dirname(__file__), "data/test11.jpeg")
|
166 |
test12 = os.path.join(os.path.dirname(__file__), "data/test12.jpeg")
|
167 |
|
168 |
+
base_image = gr.Interface(
|
169 |
fn=predict,
|
170 |
inputs=gr.Image(type="pil"),
|
171 |
outputs=gr.Image(type="pil"),
|
172 |
+
title="Luffy and Chopper face detection (Base mobile net model)",
|
173 |
+
description="Upload a Image for prediction or click on below examples",
|
174 |
examples=[[test1],[test2],[test3],[test4],[test5],[test6],[test7],[test8],[test9],[test10],[test11],[test12],],
|
175 |
cache_examples=True
|
176 |
)#.launch(share=True)
|
177 |
|
178 |
+
tuned_image = gr.Interface(
|
179 |
+
fn=predict3,
|
180 |
+
inputs=gr.Image(type="pil"),
|
181 |
+
outputs=gr.Image(type="pil"),
|
182 |
+
title="Luffy and Chopper face detection (tuned mobile net model)",
|
183 |
+
description="Upload a Image for prediction or click on below examples. Mobile net tuned with data Augmentation",
|
184 |
+
examples=[[test1],[test2],[test3],[test4],[test5],[test6],[test7],[test8],[test9],[test10],[test11],[test12],],
|
185 |
+
cache_examples=True
|
186 |
+
)#.launch(share=True)
|
187 |
+
|
188 |
+
|
189 |
+
|
190 |
a = os.path.join(os.path.dirname(__file__), "data/a.mp4") # Video
|
191 |
b = os.path.join(os.path.dirname(__file__), "data/b.mp4") # Video
|
192 |
c = os.path.join(os.path.dirname(__file__), "data/c.mp4") # Video
|
193 |
|
|
|
194 |
video_out_file = os.path.join(samples_folder,'detected' + '.mp4')
|
195 |
|
196 |
stt_demo = gr.Interface(
|
|
|
205 |
cache_examples=False
|
206 |
)
|
207 |
|
208 |
+
demo = gr.TabbedInterface([base_image,tuned_image, stt_demo], ["Image (base model)","Image (tuned model)", "Video"])
|
209 |
|
210 |
if __name__ == "__main__":
|
211 |
demo.launch()
|