def bbox_pipeline(bbox, img, bbox_list=[]):
'''
Processing vehicle detection and bounding box.
'''
img = np.copy(img)
# Do multi-scale searching
scale = 1.0
bbox_list = bbox.find_cars(img, scale, bbox_list)
scale = 1.5
bbox_list = bbox.find_cars(img, scale, bbox_list)
scale = 2.0
bbox_list = bbox.find_cars(img, scale, bbox_list)
### Heatmap and labelledbounding box
# Heat map
heat = np.zeros_like(img[:, :, 0]).astype(np.float)
# Add heat to each box in box list
heat = bbox.add_heat(heat, bbox_list)
# Apply threshold to help remove false positives
heat = bbox.apply_threshold(heat, 10)
# Visualize the heatmap when displaying
heatmap = np.clip(heat, 0, 255)
# Label bounding box
# Find final boxes from heatmap using label function
labels = label(heatmap)
# img = bbox.draw_bboxes(img, bbox_list)
# draw_img = bbox.draw_labeled_bboxes(img, labels)
# To view the heatmap boxes?
draw_img = np.array(np.dstack((heatmap, heatmap, heatmap)) * 255,
dtype='uint8')
# Alpha blending
draw_img = cv2.addWeighted(draw_img, 0.5, img, 0.5, 0)
# Searching window (big and small)
s_win = ((bbox.xstart_s, bbox.ystart_s), (bbox.xstop_s, bbox.ystop_s))
b_win = ((bbox.xstart, bbox.ystart), (bbox.xstop, bbox.ystop))
cv2.rectangle(draw_img, s_win[0], s_win[1], (255, 0, 0), 2)
cv2.rectangle(draw_img, b_win[0], b_win[1], (255, 0, 0), 2)
return draw_img
scale = 2.0 bbox_list = bbox.find_cars(img, scale, bbox_list) scale = 2.2 bbox_list = bbox.find_cars(img, scale, bbox_list) ### Heatmap and labelledbounding box # Heat map heat = np.zeros_like(img[:, :, 0]).astype(np.float) # Add heat to each box in box list heat = bbox.add_heat(heat, bbox_list) # Apply threshold to help remove false positives heat = bbox.apply_threshold(heat, 0) # Visualize the heatmap when displaying heatmap = np.clip(heat, 0, 255) # Label bounding box # Find final boxes from heatmap using label function labels = label(heatmap) draw_img = bbox.draw_labeled_bboxes(np.copy(img), labels) # Show heat map plt.imshow(heatmap) plt.show() # Show labelled image plt.imshow(draw_img)