I am trying webcam_od.py example. The frame rate is very slow, how to increase it?
Hi, the code originally stems from here, with a few updates: roboflow-api-snippets/Python/webcam at main · roboflow/roboflow-api-snippets · GitHub
You can try resizing the images prior to sending them for inference to increase the inference speed. A “smaller” image means less data being transferred, so the overall prediction can process faster.
I’ve already added this to the webcam_od.py, just be sure you’ve set the ROBOFLOW_SIZE parameter in your roboflow_config.json file - roboflow-computer-vision-utilities/roboflow_config.json at main · roboflow/roboflow-computer-vision-utilities · GitHub
- line 14 of
webcam_od.pycontainsROBOFLOW_SIZE = config["ROBOFLOW_SIZE"] - lines 43-46 of
webcam_od.pycontains:
# Resize (while maintaining the aspect ratio) to improve speed and save bandwidth
height, width, channels = img.shape
scale = ROBOFLOW_SIZE / max(height, width)
img = cv2.resize(img, (round(scale * width), round(scale * height)))
ROBOFLOW_SIZE would come from the value you resized your images to with the Resize preprocessing step during version generation.
- More on image preprocessing and the Resize preprocessing feature: Image Preprocessing - Roboflow
- More on Resizing images for training of computer vision models: You Might Be Resizing Your Images Incorrectly