Spaces:
Sleeping
Sleeping
Update my_model/object_detection.py
Browse files- my_model/object_detection.py +4 -11
my_model/object_detection.py
CHANGED
@@ -8,8 +8,6 @@ import os
|
|
8 |
from my_model.gen_utilities import get_image_path, get_model_path ,show_image
|
9 |
|
10 |
|
11 |
-
|
12 |
-
|
13 |
class ObjectDetector:
|
14 |
"""
|
15 |
A class for detecting objects in images using models like Detic and YOLOv5.
|
@@ -63,7 +61,6 @@ class ObjectDetector:
|
|
63 |
|
64 |
try:
|
65 |
model_path = get_model_path('deformable-detr-detic')
|
66 |
-
|
67 |
self.processor = AutoImageProcessor.from_pretrained(model_path)
|
68 |
self.model = AutoModelForObjectDetection.from_pretrained(model_path)
|
69 |
except Exception as e:
|
@@ -115,8 +112,7 @@ class ObjectDetector:
|
|
115 |
print(f"Error processing image: {e}")
|
116 |
raise
|
117 |
|
118 |
-
|
119 |
-
|
120 |
def detect_objects(self, image, threshold=0.4):
|
121 |
"""
|
122 |
Detect objects in the given image using the loaded model.
|
@@ -139,6 +135,7 @@ class ObjectDetector:
|
|
139 |
else:
|
140 |
raise ValueError("Model not loaded or unsupported model name")
|
141 |
|
|
|
142 |
def _detect_with_detic(self, image, threshold):
|
143 |
"""
|
144 |
Detect objects using the Detic model.
|
@@ -155,9 +152,7 @@ class ObjectDetector:
|
|
155 |
inputs = self.processor(images=image, return_tensors="pt")
|
156 |
outputs = self.model(**inputs)
|
157 |
target_sizes = torch.tensor([image.size[::-1]])
|
158 |
-
results = self.processor.post_process_object_detection(outputs, target_sizes=target_sizes, threshold=threshold)[
|
159 |
-
0]
|
160 |
-
|
161 |
detected_objects_str = ""
|
162 |
detected_objects_list = []
|
163 |
for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
|
@@ -169,6 +164,7 @@ class ObjectDetector:
|
|
169 |
detected_objects_list.append((label_name, box_rounded, certainty))
|
170 |
return detected_objects_str, detected_objects_list
|
171 |
|
|
|
172 |
def _detect_with_yolov5(self, image, threshold):
|
173 |
"""
|
174 |
Detect objects using the YOLOv5 model.
|
@@ -184,7 +180,6 @@ class ObjectDetector:
|
|
184 |
|
185 |
cv2_img = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
|
186 |
results = self.model(cv2_img)
|
187 |
-
|
188 |
detected_objects_str = ""
|
189 |
detected_objects_list = []
|
190 |
for *bbox, conf, cls in results.xyxy[0]:
|
@@ -214,7 +209,6 @@ class ObjectDetector:
|
|
214 |
font = ImageFont.truetype("arial.ttf", 15)
|
215 |
except IOError:
|
216 |
font = ImageFont.load_default()
|
217 |
-
|
218 |
colors = ["red", "green", "blue", "yellow", "purple", "orange"]
|
219 |
label_color_map = {}
|
220 |
|
@@ -224,7 +218,6 @@ class ObjectDetector:
|
|
224 |
|
225 |
color = label_color_map[label_name]
|
226 |
draw.rectangle(box, outline=color, width=3)
|
227 |
-
|
228 |
label_text = f"{label_name}"
|
229 |
if show_confidence:
|
230 |
label_text += f" ({round(score, 2)}%)"
|
|
|
8 |
from my_model.gen_utilities import get_image_path, get_model_path ,show_image
|
9 |
|
10 |
|
|
|
|
|
11 |
class ObjectDetector:
|
12 |
"""
|
13 |
A class for detecting objects in images using models like Detic and YOLOv5.
|
|
|
61 |
|
62 |
try:
|
63 |
model_path = get_model_path('deformable-detr-detic')
|
|
|
64 |
self.processor = AutoImageProcessor.from_pretrained(model_path)
|
65 |
self.model = AutoModelForObjectDetection.from_pretrained(model_path)
|
66 |
except Exception as e:
|
|
|
112 |
print(f"Error processing image: {e}")
|
113 |
raise
|
114 |
|
115 |
+
|
|
|
116 |
def detect_objects(self, image, threshold=0.4):
|
117 |
"""
|
118 |
Detect objects in the given image using the loaded model.
|
|
|
135 |
else:
|
136 |
raise ValueError("Model not loaded or unsupported model name")
|
137 |
|
138 |
+
|
139 |
def _detect_with_detic(self, image, threshold):
|
140 |
"""
|
141 |
Detect objects using the Detic model.
|
|
|
152 |
inputs = self.processor(images=image, return_tensors="pt")
|
153 |
outputs = self.model(**inputs)
|
154 |
target_sizes = torch.tensor([image.size[::-1]])
|
155 |
+
results = self.processor.post_process_object_detection(outputs, target_sizes=target_sizes, threshold=threshold)[0]
|
|
|
|
|
156 |
detected_objects_str = ""
|
157 |
detected_objects_list = []
|
158 |
for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
|
|
|
164 |
detected_objects_list.append((label_name, box_rounded, certainty))
|
165 |
return detected_objects_str, detected_objects_list
|
166 |
|
167 |
+
|
168 |
def _detect_with_yolov5(self, image, threshold):
|
169 |
"""
|
170 |
Detect objects using the YOLOv5 model.
|
|
|
180 |
|
181 |
cv2_img = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
|
182 |
results = self.model(cv2_img)
|
|
|
183 |
detected_objects_str = ""
|
184 |
detected_objects_list = []
|
185 |
for *bbox, conf, cls in results.xyxy[0]:
|
|
|
209 |
font = ImageFont.truetype("arial.ttf", 15)
|
210 |
except IOError:
|
211 |
font = ImageFont.load_default()
|
|
|
212 |
colors = ["red", "green", "blue", "yellow", "purple", "orange"]
|
213 |
label_color_map = {}
|
214 |
|
|
|
218 |
|
219 |
color = label_color_map[label_name]
|
220 |
draw.rectangle(box, outline=color, width=3)
|
|
|
221 |
label_text = f"{label_name}"
|
222 |
if show_confidence:
|
223 |
label_text += f" ({round(score, 2)}%)"
|