File size: 9,999 Bytes
17c1e65 311b9c4 f0f6ecf 17c1e65 311b9c4 17c1e65 4084ed7 17c1e65 4084ed7 17c1e65 4084ed7 17c1e65 4084ed7 7927b9f 4084ed7 7927b9f 17c1e65 4084ed7 7927b9f 17c1e65 08b4416 17c1e65 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 |
from transformers import AutoImageProcessor, AutoModelForObjectDetection
import torch
from PIL import Image, ImageDraw, ImageFont
import numpy as np
import cv2
import os
from my_model.utilities import get_image_path, get_model_path ,show_image
class ObjectDetector:
"""
A class for detecting objects in images using models like Detic and YOLOv5.
This class supports loading and using different object detection models to identify objects
in images and draw bounding boxes around them.
Attributes:
model (torch.nn.Module): The loaded object detection model.
processor (transformers.AutoImageProcessor): Processor for the Detic model.
model_name (str): Name of the model used for detection.
"""
def __init__(self):
"""
Initializes the ObjectDetector class with default values.
"""
self.model = None
self.processor = None
self.model_name = None
def load_model(self, model_name='detic', pretrained=True, model_version='yolov5s'):
"""
Load the specified object detection model.
Args:
model_name (str): Name of the model to load. Options are 'detic' and 'yolov5'.
pretrained (bool): Boolean indicating if a pretrained model should be used.
model_version (str): Version of the YOLOv5 model, applicable only when using YOLOv5.
Raises:
ValueError: If an unsupported model name is provided.
"""
self.model_name = model_name
if model_name == 'detic':
self._load_detic_model(pretrained)
elif model_name == 'yolov5':
self._load_yolov5_model(pretrained, model_version)
else:
raise ValueError(f"Unsupported model name: {model_name}")
def _load_detic_model(self, pretrained):
"""
Load the Detic model.
Args:
pretrained (bool): If True, load a pretrained model.
"""
try:
model_path = get_model_path('deformable-detr-detic')
self.processor = AutoImageProcessor.from_pretrained(model_path)
self.model = AutoModelForObjectDetection.from_pretrained(model_path)
except Exception as e:
print(f"Error loading Detic model: {e}")
raise
def _load_yolov5_model(self, pretrained, model_version):
"""
Load the YOLOv5 model.
Args:
pretrained (bool): If True, load a pretrained model.
model_version (str): Version of the YOLOv5 model.
"""
try:
model_path = get_path('yolov5', 'models')
if model_path and os.path.exists(model_path):
self.model = torch.hub.load(model_path, model_version, pretrained=pretrained, source='local')
else:
self.model = torch.hub.load('ultralytics/yolov5', model_version, pretrained=pretrained)
except Exception as e:
print(f"Error loading YOLOv5 model: {e}")
raise
def process_image(self, image_input):
"""
Process the image from the given path or file-like object.
Args:
image_input (str or file-like object): Path to the image file or a file-like object.
Returns:
Image.Image: Processed image in RGB format.
Raises:
Exception: If an error occurs during image processing.
"""
try:
# Check if the input is a string (path) or a file-like object
if isinstance(image_input, str):
# Open the image from a file path
with Image.open(image_input) as image:
return image.convert("RGB")
elif hasattr(image_input, 'read'):
# If image_input is a file-like object, open it as an image
return Image.open(image_input).convert("RGB")
else:
# If image_input is already a PIL Image, just convert it
return image_input.convert("RGB")
except Exception as e:
print(f"Error processing image: {e}")
raise
def detect_objects(self, image, threshold=0.4):
"""
Detect objects in the given image using the loaded model.
Args:
image (Image.Image): Image in which to detect objects.
threshold (float): Model detection confidence.
Returns:
tuple: A tuple containing a string representation and a list of detected objects.
Raises:
ValueError: If the model is not loaded or the model name is unsupported.
"""
if self.model_name == 'detic':
return self._detect_with_detic(image, threshold)
elif self.model_name == 'yolov5':
return self._detect_with_yolov5(image, threshold)
else:
raise ValueError("Model not loaded or unsupported model name")
def _detect_with_detic(self, image, threshold):
"""
Detect objects using the Detic model.
Args:
image (Image.Image): The image in which to detect objects.
threshold (float): The confidence threshold for detections.
Returns:
tuple: A tuple containing a string representation and a list of detected objects.
Each object in the list is represented as a tuple (label_name, box_rounded, certainty).
"""
inputs = self.processor(images=image, return_tensors="pt")
outputs = self.model(**inputs)
target_sizes = torch.tensor([image.size[::-1]])
results = self.processor.post_process_object_detection(outputs, target_sizes=target_sizes, threshold=threshold)[
0]
detected_objects_str = ""
detected_objects_list = []
for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
if score >= threshold:
label_name = self.model.config.id2label[label.item()]
box_rounded = [round(coord, 2) for coord in box.tolist()]
certainty = round(score.item() * 100, 2)
detected_objects_str += f"{{object: {label_name}, bounding box: {box_rounded}, certainty: {certainty}%}}\n"
detected_objects_list.append((label_name, box_rounded, certainty))
return detected_objects_str, detected_objects_list
def _detect_with_yolov5(self, image, threshold):
"""
Detect objects using the YOLOv5 model.
Args:
image (Image.Image): The image in which to detect objects.
threshold (float): The confidence threshold for detections.
Returns:
tuple: A tuple containing a string representation and a list of detected objects.
Each object in the list is represented as a tuple (label_name, box_rounded, certainty).
"""
cv2_img = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
results = self.model(cv2_img)
detected_objects_str = ""
detected_objects_list = []
for *bbox, conf, cls in results.xyxy[0]:
if conf >= threshold:
label_name = results.names[int(cls)]
box_rounded = [round(coord.item(), 2) for coord in bbox]
certainty = round(conf.item() * 100, 2)
detected_objects_str += f"{{object: {label_name}, bounding box: {box_rounded}, certainty: {certainty}%}}\n"
detected_objects_list.append((label_name, box_rounded, certainty))
return detected_objects_str, detected_objects_list
def draw_boxes(self, image, detected_objects, show_confidence=True):
"""
Draw bounding boxes around detected objects in the image.
Args:
image (Image.Image): Image on which to draw.
detected_objects (list): List of detected objects.
show_confidence (bool): Whether to show confidence scores.
Returns:
Image.Image: Image with drawn boxes.
"""
draw = ImageDraw.Draw(image)
try:
font = ImageFont.truetype("arial.ttf", 15)
except IOError:
font = ImageFont.load_default()
colors = ["red", "green", "blue", "yellow", "purple", "orange"]
label_color_map = {}
for label_name, box, score in detected_objects:
if label_name not in label_color_map:
label_color_map[label_name] = colors[len(label_color_map) % len(colors)]
color = label_color_map[label_name]
draw.rectangle(box, outline=color, width=3)
label_text = f"{label_name}"
if show_confidence:
label_text += f" ({round(score, 2)}%)"
draw.text((box[0], box[1]), label_text, fill=color, font=font)
return image
def detect_and_draw_objects(image_path, model_type='yolov5', threshold=0.2, show_confidence=True):
"""
Detects objects in an image, draws bounding boxes around them, and returns the processed image and a string description.
Args:
image_path (str): Path to the image file.
model_type (str): Type of model to use for detection ('yolov5' or 'detic').
threshold (float): Detection threshold.
show_confidence (bool): Whether to show confidence scores on the output image.
Returns:
tuple: A tuple containing the processed Image.Image and a string of detected objects.
"""
detector = ObjectDetector()
detector.load_model(model_type)
image = detector.process_image(image_path)
detected_objects_string, detected_objects_list = detector.detect_objects(image, threshold=threshold)
image_with_boxes = detector.draw_boxes(image, detected_objects_list, show_confidence=show_confidence)
return image_with_boxes, detected_objects_string
if __name__ == "__main__":
pass
|