[Python] How to improve this class to have reduced memory usage and better performance?

I'm trying to increase my FPS of my program, but this class is using too much memory which slows down performance. How can I improve this class so my program can take screenshots faster and perform better.

import cv2
import numpy as np

# Work on this to cut down memory and boost performance
class ObjectDetection:

def __init__(self):
self.avg_y_obstacle = None
self.avg_x_obstacle = None
self.avg_y_jump = None
self.avg_x_jump = None

# Preload templates for obstacle detection to optimize performance
self.templates = {
'template1': cv2.imread('template1.png', 0),
'template2': cv2.imread('template2.png', 0),
'template3': cv2.imread('template3.png', 0),
'template4': cv2.imread('template4.png', 0),
'template5': cv2.imread('template5.png', 0),
'template6': cv2.imread('template6.png', 0),
'template7': cv2.imread('template7.png', 0),
'template8': cv2.imread('template8.png', 0)
}
# Check if all templates loaded properly
for name, template in self.templates.items():
if template is None:
raise ValueError(f"Error loading template: {name}")

def ObstacleDetect(self, imageInput=None, threshold=0.8):
if imageInput is None:
raise ValueError("Input image is required")

# Convert input image to grayscale
image_gray = cv2.cvtColor(imageInput, cv2.COLOR_BGR2GRAY)

coordinates = []

# Iterate over each preloaded template
for name, template in self.templates.items():
h, w = template.shape[:2] # Get template dimensions

# Perform template matching
result = cv2.matchTemplate(image_gray, template, cv2.TM_CCOEFF_NORMED)
loc = np.where(result >= threshold)

# Collect coordinates for detected objects and draw bounding boxes
for pt in zip(*loc[::-1]):
coordinates.append((pt[0], pt[1], pt[0] + w, pt[1] + h))
# cv2.rectangle(imageInput, (pt[0], pt[1]), (pt[0] + w, pt[1] + h), (0, 255, 0), 2)
# cv2.putText(imageInput, name, (pt[0], pt[1] - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 1)

# Calculate avg_x_obstacle and avg_y_obstacle if coordinates are found
if coordinates:
coordinates = np.array(coordinates)
self.avg_y_obstacle = np.mean(coordinates[:, 1]) # Average y-coordinate of top-left
self.avg_x_obstacle = np.mean(coordinates[:, 0]) # Average x-coordinate of top-left
else:
print("No obstacles detected.")
self.avg_y_obstacle = None
self.avg_x_obstacle = None

return imageInput

def findJumpArea(self, image=None, threshold=0.8):
# Load the jump template once
template = cv2.imread('jump_template.png', 0)
if image is None or template is None:
raise ValueError("Error: Could not load input image or template image.")

# Convert image to grayscale
image_gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
h, w = template.shape[:2]

# Perform template matching
result = cv2.matchTemplate(image_gray, template, cv2.TM_CCOEFF_NORMED)
loc = np.where(result >= threshold)

coordinates = []

# Collect detected jump area coordinates and draw lines
for pt in zip(*loc[::-1]):
coordinates.append((pt[0], pt[1]))
# line_length = 50
# cv2.line(image, (pt[0], pt[1]), (pt[0] + line_length, pt[1]), (0, 255, 0), 2)

# Calculate avg_x_jump and avg_y_jump if coordinates are found
if coordinates:
coordinates = np.array(coordinates)
self.avg_y_jump = np.mean(coordinates[:, 1]) # Average y-coordinate of the top-left point
self.avg_x_jump = np.mean(coordinates[:, 0]) # Average x-coordinate of the top-left point
else:
print("No jump area detected.")
self.avg_y_jump = None
self.avg_x_jump = None

return image

# Getters for average coordinates
def getObstacleAvgY(self):
return self.avg_y_obstacle

def getObstacleAvgX(self):
return self.avg_x_obstacle

def getJumpAvgY(self):
return self.avg_y_jump

def getJumpAvgX(self):
return self.avg_x_jump



I tried preloading the templates and commented out some cv visual stuff, it slightly improved performance. However, that was all I could come up with.

Continue reading...