v1

app.py

@@ -1,16 +1,21 @@
 import gradio as gr
-import numpy as np
+from transformers import pipeline
+
 
 TITLE = 'FaVQA - Fashion-related Visual Question Answering'
 DESCRIPTION = 'Fine-tuned ViLT with DeepFashion dataset images for VQA task'
-IMG_EX = np.zeros((100, 100), dtype=np.uint8)
 ARTICLE = 'COLOCAR DESCRIÇÃO AQUI MONA'
 
 
 ##########################################################
 #                 WRAPPER FUNCTION
 ##########################################################
+def predict(txt, img):
+  pipe = pipeline(model='yanka9/vilt_finetuned_deepfashionVQA_v2')
+  output = pipe(image=img, question=txt, top_k=2)
+  result = {item['answer']:item['score'] for item in output}
 
+  return result
 
 ##########################################################
 #                  COMPONENTS
@@ -22,27 +27,27 @@ txt_input = gr.Textbox(
 )
 
 img_input = gr.Image(
-  value=IMG_EX,
-  sources=['upload', 'clipboard']
+  sources=['upload', 'clipboard'],
+  type='pil'
 )
 
-txt_output = gr.Textbox(
-  max_lines=2,
+label_output = gr.Label(
   label='Answer',
+  num_top_classes=2
 )
 
 ##########################################################
 #                  BUILDER
 ##########################################################
 demo = gr.Interface(
-  fn=gr.load('models/yanka9/vilt_finetuned_deepfashionVQA_v2'),
+  fn=predict,
   title=TITLE,
   description=DESCRIPTION,
   article=ARTICLE,
   theme='gradio/monochrome',
   allow_flagging='never',
   inputs=[txt_input, img_input],
-  outputs=txt_output
+  outputs=label_output
 )
 
 demo.launch()