Spaces:
Running
Running
image to story
Browse files
app.py
ADDED
@@ -0,0 +1,87 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import pipeline
|
2 |
+
from langchain_core.runnables import RunnableLambda
|
3 |
+
from langchain_huggingface import HuggingFacePipeline
|
4 |
+
from PIL import Image
|
5 |
+
|
6 |
+
pipe1 = pipeline("object-detection", model="facebook/detr-resnet-50")
|
7 |
+
pipe2 = pipeline("image-to-text", model="nlpconnect/vit-gpt2-image-captioning")
|
8 |
+
repo_id = "mistralai/Mistral-7B-Instruct-v0.2"
|
9 |
+
|
10 |
+
|
11 |
+
llm = HuggingFacePipeline.from_model_id(
|
12 |
+
model_id=repo_id,
|
13 |
+
task="text-generation",
|
14 |
+
pipeline_kwargs={"max_length": 100,"temperature":0.5},
|
15 |
+
)
|
16 |
+
def reduce_add(a):
|
17 |
+
ll=dict()
|
18 |
+
for i in a:
|
19 |
+
if i['score'] > 0.89:
|
20 |
+
if i['label'] not in ll.keys():
|
21 |
+
ll[i['label']] = 1
|
22 |
+
else:
|
23 |
+
ll[i['label']]+=1
|
24 |
+
return "there are \n"+', \n'.join([str(i[1])+' '+i[0] for i in ll.items() ])
|
25 |
+
|
26 |
+
def image_segmentation_tool(image: str):
|
27 |
+
# image = Image.open(image_path)
|
28 |
+
segmentation_results = pipe1(image)
|
29 |
+
if reduce_add(segmentation_results) == "there are \n":
|
30 |
+
raise Passs()
|
31 |
+
return reduce_add(segmentation_results)
|
32 |
+
|
33 |
+
def image_caption_tool(image: str):
|
34 |
+
# image = Image.open(image_path)
|
35 |
+
segmentation_results = pipe2(image)
|
36 |
+
if segmentation_results[0]["generated_text"] == "":
|
37 |
+
raise Passs("no result found use different image to create story")
|
38 |
+
return segmentation_results[0]["generated_text"]
|
39 |
+
|
40 |
+
from langchain_core.prompts import PromptTemplate
|
41 |
+
|
42 |
+
|
43 |
+
def story_generation_tool(segmentation_results):
|
44 |
+
prompt_template = """
|
45 |
+
You are a storyteller. Based on the following segmentation results, create a story:
|
46 |
+
{segmentation_results}
|
47 |
+
|
48 |
+
Story:
|
49 |
+
"""
|
50 |
+
prompt = PromptTemplate.from_template(prompt_template)
|
51 |
+
story = prompt | llm
|
52 |
+
return story.invoke(input={"segmentation_results":segmentation_results})
|
53 |
+
|
54 |
+
# def translation_tool(english_text):
|
55 |
+
# prompt_template = """
|
56 |
+
# You are a translator. Translate the following English text to Hindi:
|
57 |
+
# {english_text}
|
58 |
+
|
59 |
+
# Translation:
|
60 |
+
# """
|
61 |
+
# prompt = PromptTemplate.from_template(prompt_template)
|
62 |
+
# translation = prompt | llm
|
63 |
+
# return translation.invoke(input={"english_text": english_text})
|
64 |
+
|
65 |
+
|
66 |
+
runnable = RunnableLambda(image_segmentation_tool).with_fallbacks([RunnableLambda(image_caption_tool)])
|
67 |
+
runnable2 = RunnableLambda(story_generation_tool)
|
68 |
+
# runnable3 = RunnableLambda(translation_tool)
|
69 |
+
|
70 |
+
chain = runnable | runnable2
|
71 |
+
|
72 |
+
import gradio as gr
|
73 |
+
|
74 |
+
title = "Image to short Story Generator"
|
75 |
+
description = """
|
76 |
+
Upload an image, and this app will generate a short story based on the image.
|
77 |
+
"""
|
78 |
+
|
79 |
+
def sepia(input_img):
|
80 |
+
sepia_img=chain.invoke(input_img)
|
81 |
+
return sepia_img
|
82 |
+
|
83 |
+
demo = gr.Interface(sepia, gr.Image(type='pil'), "textarea",title=title,
|
84 |
+
description=description,live=True
|
85 |
+
)
|
86 |
+
if __name__ == "__main__":
|
87 |
+
demo.launch()
|