File size: 6,263 Bytes
c11ae87 | 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 | """Attend DeepFace demo — free HF Space (CPU).
Slim surface for portfolio demos: detect + Facenet512 embed + in-memory gallery.
Not the full FastAPI attendance product (auth/schedules/WS liveness).
"""
from __future__ import annotations
import json
import tempfile
from pathlib import Path
import cv2
import gradio as gr
import numpy as np
from deepface import DeepFace
MODEL_NAME = "Facenet512"
DETECTOR = "opencv"
THRESHOLD = 0.45
# Ephemeral gallery for this Space process (clears on restart/sleep).
_gallery: list[dict] = []
def _cosine(a: np.ndarray, b: np.ndarray) -> float:
return float(np.dot(a, b) / (np.linalg.norm(a) * np.linalg.norm(b) + 1e-10))
def _represent(path: str) -> list[dict]:
return DeepFace.represent(
img_path=path,
model_name=MODEL_NAME,
enforce_detection=True,
detector_backend=DETECTOR,
anti_spoofing=False,
)
def _save_upload(image: np.ndarray) -> str:
if image is None:
raise gr.Error("Upload an image first.")
# Gradio gives RGB; OpenCV/DeepFace file path is fine as PNG.
rgb = image if image.ndim == 3 else cv2.cvtColor(image, cv2.COLOR_GRAY2RGB)
fd, path = tempfile.mkstemp(suffix=".png")
Path(path).write_bytes(b"") # ensure file exists for cv2
import os
os.close(fd)
# Write via cv2 expecting BGR
bgr = cv2.cvtColor(rgb.astype(np.uint8), cv2.COLOR_RGB2BGR)
if not cv2.imwrite(path, bgr):
raise gr.Error("Could not write temp image.")
return path
def detect_faces(image: np.ndarray) -> tuple[np.ndarray | None, str]:
path = _save_upload(image)
try:
faces = _represent(path)
except Exception as exc: # DeepFace raises broadly
return image, json.dumps({"ok": False, "error": str(exc)}, indent=2)
finally:
Path(path).unlink(missing_ok=True)
annotated = image.copy()
rows = []
for i, face in enumerate(faces):
area = face.get("facial_area") or {}
x, y, w, h = int(area.get("x", 0)), int(area.get("y", 0)), int(area.get("w", 0)), int(area.get("h", 0))
if w and h:
cv2.rectangle(annotated, (x, y), (x + w, y + h), (0, 180, 120), 2)
cv2.putText(annotated, f"face-{i}", (x, max(0, y - 8)), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 180, 120), 2)
emb = np.array(face["embedding"], dtype=float)
rows.append(
{
"index": i,
"facial_area": area,
"embedding_dim": int(emb.shape[0]),
"embedding_preview": emb[:8].round(4).tolist(),
}
)
return annotated, json.dumps({"ok": True, "faces": rows, "model": MODEL_NAME}, indent=2)
def enroll(image: np.ndarray, name: str) -> str:
name = (name or "").strip() or f"person-{len(_gallery) + 1}"
path = _save_upload(image)
try:
faces = _represent(path)
except Exception as exc:
return json.dumps({"ok": False, "error": str(exc)}, indent=2)
finally:
Path(path).unlink(missing_ok=True)
if not faces:
return json.dumps({"ok": False, "error": "No face detected"}, indent=2)
emb = np.array(faces[0]["embedding"], dtype=float)
_gallery.append({"name": name, "embedding": emb})
return json.dumps({"ok": True, "enrolled": name, "gallery_size": len(_gallery)}, indent=2)
def recognize(image: np.ndarray) -> tuple[np.ndarray | None, str]:
if not _gallery:
return image, json.dumps({"ok": False, "error": "Gallery empty — enroll someone first."}, indent=2)
path = _save_upload(image)
try:
faces = _represent(path)
except Exception as exc:
return image, json.dumps({"ok": False, "error": str(exc)}, indent=2)
finally:
Path(path).unlink(missing_ok=True)
annotated = image.copy()
results = []
for i, face in enumerate(faces):
emb = np.array(face["embedding"], dtype=float)
best_name, best = "Unknown", -1.0
for item in _gallery:
score = _cosine(emb, item["embedding"])
if score > best:
best, best_name = score, item["name"]
matched = best >= THRESHOLD
label = f"{best_name} ({best:.2f})" if matched else f"Unknown ({best:.2f})"
area = face.get("facial_area") or {}
x, y, w, h = int(area.get("x", 0)), int(area.get("y", 0)), int(area.get("w", 0)), int(area.get("h", 0))
color = (0, 180, 120) if matched else (200, 80, 60)
if w and h:
cv2.rectangle(annotated, (x, y), (x + w, y + h), color, 2)
cv2.putText(annotated, label, (x, max(0, y - 8)), cv2.FONT_HERSHEY_SIMPLEX, 0.55, color, 2)
results.append({"index": i, "name": best_name if matched else "Unknown", "score": round(best, 4), "matched": matched})
return annotated, json.dumps({"ok": True, "results": results, "threshold": THRESHOLD}, indent=2)
def clear_gallery() -> str:
_gallery.clear()
return json.dumps({"ok": True, "gallery_size": 0}, indent=2)
with gr.Blocks(title="Attend DeepFace (CPU)") as demo:
gr.Markdown(
"""# Attend — DeepFace demo (free CPU)
Upload a face photo to detect embeddings (Facenet512) or enroll/recognize against an ephemeral gallery.
Full attendance API (schedules, liveness WS, SQLite) is **not** hosted here — see repo + Azure FE shell."""
)
with gr.Tab("Detect"):
inp = gr.Image(type="numpy", label="Face image")
out_img = gr.Image(type="numpy", label="Annotated")
out_json = gr.Code(language="json", label="Result")
gr.Button("Detect").click(detect_faces, inputs=inp, outputs=[out_img, out_json])
with gr.Tab("Enroll / Recognize"):
name = gr.Textbox(label="Name for enroll", placeholder="Alice")
inp2 = gr.Image(type="numpy", label="Face image")
out_img2 = gr.Image(type="numpy", label="Annotated")
out_json2 = gr.Code(language="json", label="Result")
with gr.Row():
gr.Button("Enroll").click(enroll, inputs=[inp2, name], outputs=out_json2)
gr.Button("Recognize").click(recognize, inputs=inp2, outputs=[out_img2, out_json2])
gr.Button("Clear gallery").click(clear_gallery, outputs=out_json2)
if __name__ == "__main__":
demo.queue(max_size=8).launch()
|