Robot Perception
Prof Muthukumaran K
Vishvender Tyagi
21BRS1203
Lab RCNN Object Detection
Code
import cv2
import numpy as np
from sklearn.svm import SVC
from sklearn.preprocessing import StandardScaler
import pickle
import os
from google.colab.patches import cv2_imshow
# Placeholder function to simulate feature extraction
def extract_features(image):
if isinstance(image, np.ndarray):
return image.flatten()[:100]
else:
raise ValueError("Input to extract_features must be a numpy array
representing an image.")
def load_image_and_propose_regions(image_path):
# Check if the file exists
if not os.path.isfile(image_path):
raise FileNotFoundError(f"The image file at {image_path} was not
found.")
# Load the image
image = cv2.imread(image_path)
if image is None:
raise ValueError(f"Failed to load image from {image_path}. Please
check the file format and path.")
height, width = image.shape[:2]
# Generate region proposals (using Selective Search)
ss = cv2.ximgproc.segmentation.createSelectiveSearchSegmentation()
ss.setBaseImage(image)
ss.switchToSelectiveSearchFast()
rects = ss.process()
proposals = []
for (x, y, w, h) in rects[:100]: # Limit to 100 proposals
proposals.append((x, y, w, h, image[y:y + h, x:x + w])) # Append
the coordinates and patch
return image, proposals
# Train SVM model
def train_svm(X_train, y_train):
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
svm = SVC(probability=True)
svm.fit(X_train_scaled, y_train)
return svm, scaler
def predict(svm, scaler, proposals):
results = []
for proposal in proposals:
features = extract_features(proposal[4]) # Use the extracted
image patch
features_scaled = scaler.transform([features])
prob = svm.predict_proba(features_scaled)
results.append(prob[0]) # Get probabilities for classes
return results
def main():
# Load dataset
image_path = 'rcnn_image2.jpg' # Your input image path
image, proposals = load_image_and_propose_regions(image_path)
X_train = []
y_train = []
for i in range(len(proposals)):
features = extract_features(proposals[i][4])
X_train.append(features)
y_train.append(1 if i % 2 == 0 else 0) # Dummy labels
X_train = np.array(X_train)
y_train = np.array(y_train)
# Train SVM
svm, scaler = train_svm(X_train, y_train)
# Predict on new proposals
predictions = predict(svm, scaler, proposals)
# Draw predictions on image
for i, (x, y, w, h, _) in enumerate(proposals):
if predictions[i][1] > 0.5: # Assuming class 1 is the positive
class
cv2.rectangle(image, (x, y), (x + w, y + h), (0, 255, 0), 2)
cv2.putText(image, f"Class 1: {predictions[i][1]:.2f}", (x, y
- 5), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)
cv2_imshow(image)
cv2.waitKey(0)
cv2.destroyAllWindows()
if __name__ == "__main__":
main()
Image
Output
