load images in batch size
KnightInsight opened this issue Β· comments
Search before asking
- I have searched the YOLOv5 issues and discussions and found no similar questions.
Question
Hi, I would like to ask when using torch.hub.load to load a model, is there a method to process images in batches, accommodating scenarios where 3 or 5 images arrive simultaneously, allowing parallel execution instead of queuing?
Additional
No response
@KnightInsight hello! π
Great question! When using torch.hub.load
to load a YOLOv5 model, you can indeed process images in batches. After loading the model, you can simply pass a list of image paths or a batch of images (as a tensor) to the model's .predict()
method. This allows for parallel processing of the images, leveraging the batch processing capabilities of PyTorch and the underlying hardware acceleration (like GPUs) for efficiency.
Here's a quick example for clarity:
model = torch.hub.load('ultralytics/yolov5', 'yolov5s') # Load the model
images = ['path/to/image1.jpg', 'path/to/image2.jpg'] # List of image paths
results = model(images) # Process images in batch
Or, if you have images as tensors:
# Assuming 'imgs' is a batch of images as a tensor
results = model(imgs)
This approach should help you process multiple images simultaneously without queuing them. For more detailed information and examples, please refer to our documentation at https://docs.ultralytics.com/yolov5/.
Happy coding! π
Hi, may I know can yolov8 still able to use torch.hub.load? For example, 1 image request incoming then detect.
@KnightInsight hello π!
Yes, you can use torch.hub.load
with YOLOv8 to load the model and perform detection on incoming images. Here's a quick example:
model = torch.hub.load('ultralytics/yolov8', 'yolov8s', pretrained=True) # Load YOLOv8 model
results = model('path/to/your/image.jpg') # Perform detection on an image
results.show() # Display the results
This will load the YOLOv8 model and allow you to detect objects in a single image. For detailed documentation, please check our official docs. Happy coding! π