| import gradio as gr | |
| from fastai.vision.all import * | |
| def greet(name): | |
| return "Hello " + name + "!!" | |
| learn = load_learner('export.pkl') | |
| def pre(image): | |
| pilim= PILImageBW.create(image) | |
| t = learn.dls.test_dl([pilim], rm_type_tfms=None, num_workers=0) | |
| p,_,d= learn.get_preds(dl=t,with_decoded=True) | |
| return p.argmax().item() | |
| title = "Hand written Digit Classifier" | |
| description = " A Basic CNN trained on MNSIT Dataset using Pytorch and Fast.ai. It achieved an accuracy of 99.2 %.<br> This is my first app so i could not properly convert the image from the sketchpad to a proper 28*28 pixel image as in MNSIT database. Hence, please try to draw the digits big and in center for best accuracy. " | |
| iface = gr.Interface(fn=pre, inputs="sketchpad", outputs="text",title=title,description=description) | |
| iface.launch() |