Icevision: Pipeline for fast, single image prediction

[tugot17]

📓 New

Is this a request for a tutorial or for an example?
Example

What is the task?
How to do a fast prediction for single image

Let say we start with something like

from mantisshrimp.all import *
from PIL import Image

model = faster_rcnn.model(num_classes=5)

model.load_state_dict(torch.load(weights_path, map_location=lambda storage, loc: storage))
model.eval()

img = Image.open(path)

How to add Normalizing transformation and get a prediction without constructing the Parser and Dataset

Is this example for a specific model?
FasterRCNN

---

Don't remove
Main issue for examples: #39

[lgvaz]

We have a snippet for that, still needs to be converted to a tutorial, check it out here

[tugot17]

Yeah the approach from the example works fine. However it might be useful to add Resize tfms to the transformations list

tfms = tfms.A.Adapter([tfms.A.Resize(384, 384), tfms.A.Normalize()]) So we are sure that the image domain is same

Thanks again for your help :)

[lgvaz]

@tugot17 can this be closed?