mirror of https://github.com/huggingface/transformers.git synced 2025-07-05 05:40:05 +06:00

Add Image Processor Fast RT-DETR (#34354 )

* add fast image processor rtdetr

* add gpu/cpu test and fix docstring

* remove prints

* add to doc

* nit docstring

* avoid iterating over images/annotations several times

* change torch typing

* Add image processor fast documentation

2024-10-30 13:49:47 -04:00

3.7 KiB

Raw Blame History

Image Processor

An image processor is in charge of preparing input features for vision models and post processing their outputs. This includes transformations such as resizing, normalization, and conversion to PyTorch, TensorFlow, Flax and Numpy tensors. It may also include model specific post-processing such as converting logits to segmentation masks.

Fast image processors are available for a few models and more will be added in the future. They are based on the torchvision library and provide a significant speed-up, especially when processing on GPU. They have the same API as the base image processors and can be used as drop-in replacements. To use a fast image processor, you need to install the torchvision library, and set the use_fast argument to True when instantiating the image processor:

from transformers import AutoImageProcessor

processor = AutoImageProcessor.from_pretrained("facebook/detr-resnet-50", use_fast=True)

When using a fast image processor, you can also set the device argument to specify the device on which the processing should be done. By default, the processing is done on the same device as the inputs if the inputs are tensors, or on the CPU otherwise.

from torchvision.io import read_image
from transformers import DetrImageProcessorFast

images = read_image("image.jpg")
processor = DetrImageProcessorFast.from_pretrained("facebook/detr-resnet-50")
images_processed = processor(images, return_tensors="pt", device="cuda")

Here are some speed comparisons between the base and fast image processors for the DETR and RT-DETR models, and how they impact overall inference time: