File size: 5,119 Bytes
8dfd66b
0ecd601
 
 
bf63b36
 
0ecd601
8dfd66b
0ecd601
 
bf63b36
8dfd66b
256d97e
 
 
 
8dfd66b
0ecd601
8dfd66b
256d97e
 
 
 
8dfd66b
bf63b36
 
8dfd66b
256d97e
 
8dfd66b
256d97e
 
 
 
 
 
 
 
 
8dfd66b
256d97e
 
0ecd601
 
 
256d97e
0ecd601
8dfd66b
0ecd601
 
 
 
256d97e
8dfd66b
 
0ecd601
8dfd66b
0ecd601
 
 
256d97e
0ecd601
8dfd66b
0ecd601
8dfd66b
 
256d97e
bf63b36
256d97e
bf63b36
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8dfd66b
256d97e
 
 
0ecd601
 
8dfd66b
256d97e
8dfd66b
bf63b36
 
8dfd66b
 
bf63b36
8dfd66b
 
256d97e
bf63b36
256d97e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bf63b36
 
256d97e
 
 
 
 
 
 
 
 
 
bf63b36
256d97e
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
import io
import asyncio
import threading
import time
from fastapi import FastAPI, File, UploadFile, Header
from fastapi.responses import JSONResponse, HTMLResponse, PlainTextResponse
from PIL import Image
import torch
from transformers import AutoProcessor, AutoModelForCausalLM
import requests
import os

# ---------------------------------------------------
# FastAPI App
# ---------------------------------------------------
app = FastAPI(title="Florence Image Caption API")

device = "cuda" if torch.cuda.is_available() else "cpu"

# Lazy load model on first request (prevents HF timeout)
processor = None
model = None
model_lock = asyncio.Lock()

# Hugging Face token stored in Space secrets
HF_TOKEN = os.getenv("img2caption")

async def load_model():
    global processor, model

    if model is None:
        processor = AutoProcessor.from_pretrained(
            "microsoft/Florence-2-base",
            trust_remote_code=True
        )
        model = AutoModelForCausalLM.from_pretrained(
            "microsoft/Florence-2-base",
            trust_remote_code=True
        ).to(device).eval()


def run_caption(image: Image.Image) -> str:
    inputs = processor(
        text="<MORE_DETAILED_CAPTION>",
        images=image,
        return_tensors="pt"
    ).to(device)

    output_ids = model.generate(
        input_ids=inputs["input_ids"],
        pixel_values=inputs["pixel_values"],
        max_new_tokens=256,
        num_beams=3
    )

    decoded = processor.batch_decode(output_ids, skip_special_tokens=False)[0]

    parsed = processor.post_process_generation(
        decoded,
        task="<MORE_DETAILED_CAPTION>",
        image_size=(image.width, image.height)
    )

    return parsed["<MORE_DETAILED_CAPTION>"]


# ---------------------------------------------------
# API Endpoint (Protected only if token is sent)
# ---------------------------------------------------
@app.post("/img2caption", response_class=PlainTextResponse)
async def img2caption(
    file: UploadFile = File(...),
    authorization: str = Header(None)
):

    # If app sends a token → validate it
    if authorization is not None:
        if not authorization.startswith("Bearer "):
            return PlainTextResponse("Invalid token format", status_code=403)

        token = authorization.replace("Bearer ", "").strip()
        if token != HF_TOKEN:
            return PlainTextResponse("Invalid token", status_code=403)

    try:
        async with model_lock:
            await load_model()

        data = await file.read()
        image = Image.open(io.BytesIO(data)).convert("RGB")

        caption = run_caption(image)

        # Return ONLY the caption string, no JSON
        return caption

    except Exception as e:
        return PlainTextResponse(f"Error: {str(e)}", status_code=500)


# ---------------------------------------------------
# Simple HTML UI (no token required)
# ---------------------------------------------------
@app.get("/", response_class=HTMLResponse)
def ui():
    return """
<!DOCTYPE html>
<html>
<head>
    <title>Image Caption Generator</title>
    <style>
        body { font-family: Arial; max-width: 650px; margin: 40px auto; }
        h2 { text-align: center; }
        #preview {
            width: 100%; margin-top: 15px; display: none;
            border-radius: 8px;
        }
        #captionBox {
            margin-top: 20px; padding: 15px;
            background: #eee; border-radius: 6px; display: none;
        }
        button {
            padding: 12px; width: 100%; margin-top: 10px;
            background: #4A90E2; color: white; border: none;
            border-radius: 6px; cursor: pointer; font-size: 16px;
        }
        button:hover { background: #357ABD; }
    </style>
</head>
<body>
    <h2>Image Caption Generator</h2>
    <input type="file" id="imageInput" accept="image/*">
    <img id="preview">
    <button onclick="generateCaption()">Generate Caption</button>
    <div id="captionBox"></div>
<script>
    const imageInput = document.getElementById("imageInput");
    const preview = document.getElementById("preview");
    const captionBox = document.getElementById("captionBox");
    imageInput.onchange = () => {
        const f = imageInput.files[0];
        if (f) {
            preview.src = URL.createObjectURL(f);
            preview.style.display = "block";
        }
    };
    async function generateCaption() {
        const f = imageInput.files[0];
        if (!f) {
            alert("Upload an image first");
            return;
        }
        const form = new FormData();
        form.append("file", f);
        captionBox.style.display = "block";
        captionBox.innerHTML = "Generating caption...";
        const res = await fetch("/img2caption", {
            method: "POST",
            body: form
        });
        const text = await res.text();
        captionBox.innerHTML = text;
    }
</script>
</body>
</html>
"""


def keep_alive():
    pass


if __name__ == "__main__":
    import uvicorn
    print("🚀 Launching Fast img2caption API")
    keep_alive()
    uvicorn.run(app, host="0.0.0.0", port=7860)