Spaces:
Runtime error
Runtime error
Upload 2 files
Browse files- app.py +139 -0
- requirements.txt +3 -0
app.py
ADDED
|
@@ -0,0 +1,139 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import google.generativeai as genai
|
| 2 |
+
from google.generativeai.types import HarmBlockThreshold, HarmCategory
|
| 3 |
+
import gradio as gr
|
| 4 |
+
from PIL import Image, ImageDraw, ImageFont
|
| 5 |
+
import json
|
| 6 |
+
|
| 7 |
+
# Fetch bounding boxes and labels
|
| 8 |
+
async def get_bounding_boxes(prompt: str, image: str, api_key: str):
|
| 9 |
+
system_prompt = """
|
| 10 |
+
You are a helpful assistant, who always responds with the bounding box and label with the explanation JSON based on the user input, and nothing else.
|
| 11 |
+
Your response can also include multiple bounding boxes and their labels in the list.
|
| 12 |
+
The values in the list should be integers.
|
| 13 |
+
Here are some example responses:
|
| 14 |
+
{
|
| 15 |
+
"explanation": "User asked for the bounding box of the dragon, so I will provide the bounding box of the dragon.",
|
| 16 |
+
"bounding_boxes": [
|
| 17 |
+
{"label": "dragon", "box": [ymin, xmin, ymax, xmax]}
|
| 18 |
+
]
|
| 19 |
+
}
|
| 20 |
+
{
|
| 21 |
+
"explanation": "User asked for the bounding box of the fruits which are red in color, so I will provide the bounding box of the Apple and the Tomato.",
|
| 22 |
+
"bounding_boxes": [
|
| 23 |
+
{"label": "apple", "box": [ymin, xmin, ymax, xmax]},
|
| 24 |
+
{"label": "tomato", "box": [ymin, xmin, ymax, xmax]}
|
| 25 |
+
]
|
| 26 |
+
}
|
| 27 |
+
""".strip()
|
| 28 |
+
|
| 29 |
+
prompt = f"Return the bounding boxes and labels of: {prompt}"
|
| 30 |
+
|
| 31 |
+
messages = [
|
| 32 |
+
{"role": "user", "parts": [prompt, image]},
|
| 33 |
+
]
|
| 34 |
+
|
| 35 |
+
genai.configure(api_key=api_key)
|
| 36 |
+
|
| 37 |
+
generation_config = {
|
| 38 |
+
"temperature": 1,
|
| 39 |
+
"max_output_tokens": 8192,
|
| 40 |
+
"response_mime_type": "application/json",
|
| 41 |
+
}
|
| 42 |
+
|
| 43 |
+
model = genai.GenerativeModel(
|
| 44 |
+
model_name="gemini-1.5-flash",
|
| 45 |
+
generation_config=generation_config,
|
| 46 |
+
safety_settings={
|
| 47 |
+
HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE,
|
| 48 |
+
HarmCategory.HARM_CATEGORY_HARASSMENT: HarmBlockThreshold.BLOCK_NONE,
|
| 49 |
+
HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: HarmBlockThreshold.BLOCK_NONE,
|
| 50 |
+
HarmCategory.HARM_CATEGORY_HATE_SPEECH: HarmBlockThreshold.BLOCK_NONE
|
| 51 |
+
},
|
| 52 |
+
system_instruction=system_prompt
|
| 53 |
+
)
|
| 54 |
+
|
| 55 |
+
try:
|
| 56 |
+
response = await model.generate_content_async(messages)
|
| 57 |
+
except Exception as e:
|
| 58 |
+
if "API key not valid" in str(e):
|
| 59 |
+
raise gr.Error(
|
| 60 |
+
"Invalid API key. Please provide a valid Gemini API key.")
|
| 61 |
+
elif "rate limit" in str(e).lower():
|
| 62 |
+
raise gr.Error("Rate limit exceeded for the API key.")
|
| 63 |
+
else:
|
| 64 |
+
raise gr.Error(f"Failed to generate content: {str(e)}")
|
| 65 |
+
|
| 66 |
+
response_json = json.loads(response.text)
|
| 67 |
+
|
| 68 |
+
explanation = response_json["explanation"]
|
| 69 |
+
bounding_boxes = response_json["bounding_boxes"]
|
| 70 |
+
|
| 71 |
+
return bounding_boxes, explanation
|
| 72 |
+
|
| 73 |
+
# Adjust bounding boxes based on image size
|
| 74 |
+
async def adjust_bounding_box(bounding_boxes, image):
|
| 75 |
+
width, height = image.size
|
| 76 |
+
adjusted_boxes = []
|
| 77 |
+
for item in bounding_boxes:
|
| 78 |
+
label = item["label"]
|
| 79 |
+
ymin, xmin, ymax, xmax = [coord / 1000 for coord in item["box"]]
|
| 80 |
+
xmin *= width
|
| 81 |
+
xmax *= width
|
| 82 |
+
ymin *= height
|
| 83 |
+
ymax *= height
|
| 84 |
+
adjusted_boxes.append({"label": label, "box": [xmin, ymin, xmax, ymax]})
|
| 85 |
+
return adjusted_boxes
|
| 86 |
+
|
| 87 |
+
# Process the image and draw bounding boxes and labels
|
| 88 |
+
async def process_image(image, text, api_key):
|
| 89 |
+
if not api_key:
|
| 90 |
+
raise gr.Error("Please provide a Gemini API key.")
|
| 91 |
+
|
| 92 |
+
# Open the image using PIL
|
| 93 |
+
image = Image.open(image)
|
| 94 |
+
|
| 95 |
+
# Call the async bounding box function
|
| 96 |
+
bounding_boxes, explanation = await get_bounding_boxes(text, image, api_key)
|
| 97 |
+
|
| 98 |
+
# Adjust the bounding box based on the image dimensions
|
| 99 |
+
adjusted_boxes = await adjust_bounding_box(bounding_boxes, image)
|
| 100 |
+
|
| 101 |
+
# Draw the bounding boxes and labels on the image
|
| 102 |
+
draw = ImageDraw.Draw(image)
|
| 103 |
+
font = ImageFont.load_default(size=20)
|
| 104 |
+
|
| 105 |
+
for item in adjusted_boxes:
|
| 106 |
+
box = item["box"]
|
| 107 |
+
label = item["label"]
|
| 108 |
+
draw.rectangle(box, outline="red", width=3)
|
| 109 |
+
# Draw the label above the bounding box
|
| 110 |
+
draw.text((box[0], box[1] - 25), label, fill="red", font=font)
|
| 111 |
+
|
| 112 |
+
# Format adjusted boxes for display
|
| 113 |
+
adjusted_boxes_str = "\n".join(f"{item['label']}: {item['box']}" for item in adjusted_boxes)
|
| 114 |
+
|
| 115 |
+
return explanation, image, adjusted_boxes_str
|
| 116 |
+
|
| 117 |
+
# Gradio app
|
| 118 |
+
async def gradio_app(image, text, api_key):
|
| 119 |
+
return await process_image(image, text, api_key)
|
| 120 |
+
|
| 121 |
+
# Launch the Gradio interface
|
| 122 |
+
iface = gr.Interface(
|
| 123 |
+
fn=gradio_app,
|
| 124 |
+
inputs=[
|
| 125 |
+
gr.Image(type="filepath"),
|
| 126 |
+
gr.Textbox(label="Object(s) to detect", value="person"),
|
| 127 |
+
gr.Textbox(label="Your Gemini API Key", type="password")
|
| 128 |
+
],
|
| 129 |
+
outputs=[
|
| 130 |
+
gr.Textbox(label="Explanation"),
|
| 131 |
+
gr.Image(type="pil", label="Output Image"),
|
| 132 |
+
gr.Textbox(label="Coordinates and Labels of the Bounding Box(es)")
|
| 133 |
+
],
|
| 134 |
+
title="Gemini Object Detection ✨",
|
| 135 |
+
description="Detect objects in images using the Gemini 1.5 Flash model.",
|
| 136 |
+
allow_flagging="never"
|
| 137 |
+
)
|
| 138 |
+
|
| 139 |
+
iface.launch()
|
requirements.txt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
google-generativeai
|
| 2 |
+
gradio==5.0.0b3
|
| 3 |
+
pillow
|