Upload app.py

app.py

@@ -6,6 +6,7 @@ import cv2
 import torch
 import torch.nn as nn
 from PIL import Image
+import numpy
 import torchvision
 from torchvision import transforms
 from timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD
@@ -105,18 +106,24 @@ def run_chatbot(input, max_tokens,temperature,frequency_penalty,presence_penalty
     return conversation,[history,conversation]
 
 def run_chatbot_with_img(input_img,max_tokens,temperature,frequency_penalty,presence_penalty,gr_state=[]):
+    print(type(input_img))
+
+    img_save = Image.open(input_img.name).resize((224,224)).convert('RGB')
+    img_save.save(input_img.name)
+    img4cls = numpy.array(img_save)
     history, conversation = gr_state[0],gr_state[1]
 
     # TODO: save img and show in conversation
     # save_img(input_img)
 
 
-    img_cls = recognize_image(input_img)
+    img_cls = recognize_image(img4cls)
     # conversation = conversation+ [(f'<img src="/file={input_img.name}" style="display: inline-block;">', "")]
     input = 'I have given you a photo about '+ img_cls + ', and tell me its definition.'
     output = chat_with_GPT(input,history,max_tokens,temperature,frequency_penalty,presence_penalty)
 
-    input_mask = 'Upload image'
+    input_mask = f'<img src="/file={input_img.name}" style="display: inline-block;">'
+    # input_mask = 'Upload image'
     # conversation save chatbox content
     conversation.append((input_mask,output))
     # history for GPT
@@ -125,7 +132,7 @@ def run_chatbot_with_img(input_img,max_tokens,temperature,frequency_penalty,pres
     # chatbox gr_state
     return conversation , [history,conversation]
 
-def save_img(image):
+def save_img(image: Image.Image):
 
     filename = next(tempfile._get_candidate_names()) + '.png'
     print(filename)
@@ -151,6 +158,8 @@ def reset():
 eval_transforms = build_transforms(224)
 
 
+import openai
+import os
 
 with gr.Blocks() as demo:
     gr.HTML("""
@@ -177,7 +186,7 @@ with gr.Blocks() as demo:
         image_btn = gr.UploadButton("Upload Image", file_types=["image"])
     with gr.Column(scale=0.3, min_width=400):
             max_tokens = gr.Number(
-                minimum=500, maximum=2000, value=1000, precision=1, interactive=True, label="Maximum length of generated text")
+                value=1000, precision=1, interactive=True, label="Maximum length of generated text")
             temperature = gr.Slider(
                 minimum=0.0, maximum=1.0, value=0.0, interactive=True, label="Diversity of generated text")
             frequency_penalty = gr.Slider(minimum=-2.0, maximum=2.0, value=0.5,
@@ -196,6 +205,6 @@ with gr.Blocks() as demo:
     submit_btn.click(lambda: "", None, text_input)
     reset_btn.click(fn=reset,inputs=[],outputs=[chatbot,gr_state])
     submit_btn_img.click(run_chatbot_with_img, [image,max_tokens,temperature,frequency_penalty,presence_penalty,gr_state], [chatbot,gr_state])
-    # image_btn.upload(run_chatbot_with_img, [image,max_tokens,temperature,frequency_penalty,presence_penalty,gr_state], [chatbot,gr_state])
+    image_btn.upload(run_chatbot_with_img, [image_btn,max_tokens,temperature,frequency_penalty,presence_penalty,gr_state], [chatbot,gr_state])
 
 demo.launch(debug = True)