TorchServe default inference handlers

TorchServe provides following inference handlers out of box :

image_classifier

• Description : Handles image classification models trained on imagenet dataset.

• Input : RGB image

• Output : Top 5 predictions and their respective probability of the image

For more details refer examples

image_segmenter

• Description : Handles image segmentation models trained on imagenet dataset.

• Input : RGB image

• Output : Output shape as [CL H W], CL - number of classes, H - height and W - width.

For more details refer examples

object_detector

• Description : Handles object detection models.

• Input : RGB image

• Output : List of detected classes and bounding boxes respectively

For more details refer examples

text_classifier

• Description : Handles models trained on imagenet dataset.

• Input : text file

• Output : Class of input text

For more details refer examples