Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,68 +1,49 @@
|
|
| 1 |
import os
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
import torch
|
| 3 |
from fastapi import FastAPI, File, UploadFile
|
| 4 |
from fastapi.responses import JSONResponse, HTMLResponse
|
| 5 |
from transformers import AutoImageProcessor, AutoModelForImageClassification
|
| 6 |
from PIL import Image
|
| 7 |
|
| 8 |
-
#
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
os.environ["HF_DATASETS_CACHE"] = "/tmp/hf_cache"
|
| 12 |
-
|
| 13 |
-
# Create cache directory if missing
|
| 14 |
-
os.makedirs("/tmp/hf_cache", exist_ok=True)
|
| 15 |
-
|
| 16 |
-
# Load processor + model
|
| 17 |
-
processor = AutoImageProcessor.from_pretrained(
|
| 18 |
-
"prithivMLmods/Realistic-Gender-Classification", cache_dir="/tmp/hf_cache"
|
| 19 |
-
)
|
| 20 |
-
model = AutoModelForImageClassification.from_pretrained(
|
| 21 |
-
"prithivMLmods/Realistic-Gender-Classification", cache_dir="/tmp/hf_cache"
|
| 22 |
-
)
|
| 23 |
|
| 24 |
-
#
|
| 25 |
app = FastAPI()
|
| 26 |
|
| 27 |
@app.get("/", response_class=HTMLResponse)
|
| 28 |
async def home():
|
| 29 |
-
return
|
| 30 |
<html>
|
| 31 |
<body>
|
| 32 |
-
<h2>Upload Image for Gender Detection</h2>
|
| 33 |
<form action="/predict" enctype="multipart/form-data" method="post">
|
| 34 |
<input name="file" type="file" accept="image/*">
|
| 35 |
<input type="submit" value="Upload">
|
| 36 |
</form>
|
| 37 |
</body>
|
| 38 |
</html>
|
| 39 |
-
|
| 40 |
|
| 41 |
@app.post("/predict")
|
| 42 |
async def predict(file: UploadFile = File(...)):
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
image = Image.open(file.file).convert("RGB")
|
| 46 |
-
|
| 47 |
-
# Preprocess
|
| 48 |
-
inputs = processor(images=image, return_tensors="pt")
|
| 49 |
-
|
| 50 |
-
# Predict
|
| 51 |
-
with torch.no_grad():
|
| 52 |
-
outputs = model(**inputs)
|
| 53 |
-
probs = torch.nn.functional.softmax(outputs.logits, dim=-1)[0].cpu().numpy()
|
| 54 |
-
|
| 55 |
-
# Get labels
|
| 56 |
-
labels = list(model.config.id2label.values())
|
| 57 |
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
"male": float(probs[labels.index("male portrait")])
|
| 62 |
-
}
|
| 63 |
|
| 64 |
-
|
|
|
|
| 65 |
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
|
|
|
|
|
|
|
|
| 1 |
import os
|
| 2 |
+
os.environ["HF_HOME"] = "/tmp/hf_cache"
|
| 3 |
+
os.environ["TRANSFORMERS_CACHE"] = "/tmp/hf_cache"
|
| 4 |
+
|
| 5 |
+
import io
|
| 6 |
import torch
|
| 7 |
from fastapi import FastAPI, File, UploadFile
|
| 8 |
from fastapi.responses import JSONResponse, HTMLResponse
|
| 9 |
from transformers import AutoImageProcessor, AutoModelForImageClassification
|
| 10 |
from PIL import Image
|
| 11 |
|
| 12 |
+
# Load model and processor
|
| 13 |
+
processor = AutoImageProcessor.from_pretrained("prithivMLmods/Realistic-Gender-Classification")
|
| 14 |
+
model = AutoModelForImageClassification.from_pretrained("prithivMLmods/Realistic-Gender-Classification")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
|
| 16 |
+
# FastAPI app
|
| 17 |
app = FastAPI()
|
| 18 |
|
| 19 |
@app.get("/", response_class=HTMLResponse)
|
| 20 |
async def home():
|
| 21 |
+
return '''
|
| 22 |
<html>
|
| 23 |
<body>
|
| 24 |
+
<h2>Upload an Image for Gender Detection</h2>
|
| 25 |
<form action="/predict" enctype="multipart/form-data" method="post">
|
| 26 |
<input name="file" type="file" accept="image/*">
|
| 27 |
<input type="submit" value="Upload">
|
| 28 |
</form>
|
| 29 |
</body>
|
| 30 |
</html>
|
| 31 |
+
'''
|
| 32 |
|
| 33 |
@app.post("/predict")
|
| 34 |
async def predict(file: UploadFile = File(...)):
|
| 35 |
+
image = Image.open(io.BytesIO(await file.read())).convert("RGB")
|
| 36 |
+
inputs = processor(images=image, return_tensors="pt")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
|
| 38 |
+
with torch.no_grad():
|
| 39 |
+
logits = model(**inputs).logits
|
| 40 |
+
probs = torch.nn.functional.softmax(logits, dim=-1).cpu().numpy()[0]
|
|
|
|
|
|
|
| 41 |
|
| 42 |
+
labels = model.config.id2label
|
| 43 |
+
result = {labels[i]: float(probs[i]) for i in range(len(labels))}
|
| 44 |
|
| 45 |
+
result = {
|
| 46 |
+
"female": float(probs[0]),
|
| 47 |
+
"male": float(probs[1])
|
| 48 |
+
}
|
| 49 |
+
return JSONResponse(content=result)
|