updated model with new architecture

README.md

@@ -27,5 +27,5 @@ I randomly split the data:
 - 80% training set
 - 20% validation set
 
-After resizing the images to be 128 x 128 px, I trained using the ResNet Architecture with 18 layers and 25 epochs.
+After resizing the images to be 128 x 128 px, I trained using the ConvNeXt-22k convolution neural network  and 18 epochs.
 

app.py

@@ -21,8 +21,8 @@ def classify_image(img):
 
 learn = load_learner('dogIdentifierModel.pkl')
 
-image = gr.inputs.Image(shape=(192, 192))
-label = gr.outputs.Label()
+image = gr.components.Image(shape=(192, 192))
+label = gr.components.Label()
 examples = ['golden-retriever.jpg', 'german-shepherd.jpg', 'pinscher.jpg', 'husky.jpg']
 
 intf = gr.Interface(fn=classify_image, inputs=image, outputs=label, examples=examples)

dogIdentifierModel.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dd840bc3bb96e8d924b3354539ac89e6f4c953f40476793f62d465eab35c3715
-size 47786859
+oid sha256:6fb82faf70b0ba80b519cec9185e50891011be20ff8a7b6bd8de6072d2bc8d1d
+size 115099127

modelCreation.py

@@ -7,6 +7,7 @@ import os, shutil
 import time
 import requests
 import json
+import timm
 from azure.cognitiveservices.search.imagesearch import ImageSearchClient as api
 from msrest.authentication import CognitiveServicesCredentials as auth
 
@@ -73,7 +74,7 @@ dataloaders = DataBlock(
 ).dataloaders(path)
 
 
-learn = vision_learner(dataloaders, resnet18, metrics=error_rate)
-learn.fine_tune(25)
+learn = vision_learner(dataloaders, 'convnext_tiny_in22k', metrics=error_rate)
+learn.fine_tune(18)
 
 learn.export('dogIdentifierModel.pkl')

requirements.txt

@@ -3,4 +3,5 @@ torch
 gradio
 numpy
 pandas
+timm
 azure-cognitiveservices-search-imagesearch