initial commit

.gitattributes

@@ -1,35 +1,2 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+models/haarcascade_frontalface_alt2.xml filter=lfs diff=lfs merge=lfs -text
+models/lbfmodel.yaml filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,4 @@
+.DS_S*
+__pycache__/
+gradio_cached_examples/
+.ipynb_checkpoints/

README.md

@@ -1,12 +1,10 @@
 ---
-title: 9103H 2024F Face Align
-emoji: 🦀
-colorFrom: red
-colorTo: indigo
+title: 9103H 2024F face-align-gradio
+emoji: 🤨📐
+colorFrom: blue
+colorTo: gray
 sdk: gradio
-sdk_version: 4.44.0
+sdk_version: 4.42.0
 app_file: app.py
 pinned: false
 ---
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

align.ipynb

@@ -0,0 +1,133 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "2c2df050-8d87-467a-bac4-db82196f9476",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import cv2\n",
+    "import numpy as np\n",
+    "import matplotlib.pyplot as plt\n",
+    "\n",
+    "from math import atan2\n",
+    "from os import listdir, path\n",
+    "from PIL import Image as PImage"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "66f93746-7ea0-4856-ac86-24472166cf8c",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "img_filename = \"MT.png\"\n",
+    "\n",
+    "pimg = PImage.open(f\"./{img_filename}\").convert(\"L\")\n",
+    "pimg.thumbnail((1000,1000))\n",
+    "imgg = np.array(pimg).copy()\n",
+    "\n",
+    "iw,ih = pimg.size"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "86c17ae6-391b-474f-a157-71af179f29a0",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# https://medium.com/analytics-vidhya/facial-landmarks-and-face-detection-in-python-with-opencv-73979391f30e\n",
+    "# https://www.researchgate.net/figure/The-68-specific-human-face-landmarks_fig4_331769278\n",
+    "\n",
+    "haarcascade = \"./models/haarcascade_frontalface_alt2.xml\"\n",
+    "face_detector = cv2.CascadeClassifier(haarcascade)\n",
+    "\n",
+    "LBFmodel = \"./models/lbfmodel.yaml\"\n",
+    "landmark_detector  = cv2.face.createFacemarkLBF()\n",
+    "landmark_detector.loadModel(LBFmodel)\n",
+    "\n",
+    "faces = face_detector.detectMultiScale(imgg)\n",
+    "\n",
+    "biggest_faces = faces[np.argsort(-faces[:,2])]\n",
+    "\n",
+    "_, landmarks = landmark_detector.fit(imgg, biggest_faces)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "7724f3dd-ee63-48a3-8761-a0ae065c480f",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "OUT_W = 130\n",
+    "OUT_H = 170\n",
+    "OUT_EYE_SPACE = 60\n",
+    "OUT_NOSE_TOP = 70\n",
+    "\n",
+    "EYE_0_IDX = 36\n",
+    "EYE_1_IDX = 45\n",
+    "CHIN_IDX = 8\n",
+    "\n",
+    "for landmark in landmarks:\n",
+    "  eye0 = np.array(landmark[0][EYE_0_IDX])\n",
+    "  eye1 = np.array(landmark[0][EYE_1_IDX])\n",
+    "  chin = np.array(landmark[0][CHIN_IDX])\n",
+    "  mid = np.mean([eye0, eye1], axis=0)\n",
+    "\n",
+    "  eye_line = eye1 - eye0\n",
+    "  tilt = atan2(eye_line[1], eye_line[0])\n",
+    "  tilt_deg = 180 * tilt / np.pi\n",
+    "  print(tilt_deg)\n",
+    "\n",
+    "  chin_line = chin - mid\n",
+    "  tilt2 = atan2(chin_line[1], chin_line[0])\n",
+    "  tilt2_deg = (180 * tilt2 / np.pi) - 90\n",
+    "  print(tilt2_deg)\n",
+    "\n",
+    "  scale = OUT_EYE_SPACE / abs(eye0[0] - eye1[0])\n",
+    "\n",
+    "  # scale\n",
+    "  pimgs = pimg.resize((int(iw * scale), int(ih * scale)), resample=PImage.Resampling.LANCZOS)\n",
+    "\n",
+    "  # rotate around nose\n",
+    "  new_mid = [int(c * scale) for c in mid]\n",
+    "  crop_box = (new_mid[0] - (OUT_W // 2),\n",
+    "              new_mid[1] - OUT_NOSE_TOP,\n",
+    "              new_mid[0] + (OUT_W // 2),\n",
+    "              new_mid[1] + (OUT_H - OUT_NOSE_TOP))\n",
+    "\n",
+    "  pimgsrc0 = pimgs.rotate(0, center=new_mid, resample=PImage.Resampling.BICUBIC).crop(crop_box)\n",
+    "  display(pimgsrc0)\n",
+    "  pimgsrc = pimgs.rotate(tilt_deg, center=new_mid, resample=PImage.Resampling.BICUBIC).crop(crop_box)\n",
+    "  display(pimgsrc)\n",
+    "  pimgsrc2 = pimgs.rotate(tilt_deg, center=new_mid, resample=PImage.Resampling.BICUBIC).crop(crop_box)\n",
+    "  display(pimgsrc2)\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

app.py

@@ -0,0 +1,93 @@
+import cv2
+import gradio as gr
+import numpy as np
+import matplotlib.pyplot as plt
+
+from math import atan2
+from os import listdir, path
+from PIL import Image as PImage
+
+OUT_W = 130
+OUT_H = 170
+OUT_EYE_SPACE = 64
+OUT_NOSE_TOP = 72
+
+EYE_0_IDX = 36
+EYE_1_IDX = 45
+
+haarcascade = "./models/haarcascade_frontalface_alt2.xml"
+face_detector = cv2.CascadeClassifier(haarcascade)
+
+LBFmodel = "./models/lbfmodel.yaml"
+landmark_detector  = cv2.face.createFacemarkLBF()
+landmark_detector.loadModel(LBFmodel)
+
+NUM_OUTS = 16
+all_outputs = [gr.Image(format="jpeg") for _ in range(NUM_OUTS)]
+
+def face(img_in):
+  out_pad = NUM_OUTS * [gr.Image(visible=False)]
+  if img_in is None:
+    return out_pad
+
+  pimg = img_in.convert("L")
+  pimg.thumbnail((1000,1000))
+  imgg = np.array(pimg).copy()
+
+  iw,ih = pimg.size
+  
+  faces = face_detector.detectMultiScale(imgg)
+
+  if len(faces) < 1:
+    return out_pad
+
+  biggest_faces = faces[np.argsort(-faces[:,2])]
+  _, landmarks = landmark_detector.fit(imgg, biggest_faces)
+
+  if len(landmarks) < 1:
+    return out_pad
+
+  out_images = []
+  for landmark in landmarks:
+    eye0 = np.array(landmark[0][EYE_0_IDX])
+    eye1 = np.array(landmark[0][EYE_1_IDX])
+    mid = np.mean([eye0, eye1], axis=0)
+
+    eye_line = eye1 - eye0
+    tilt = atan2(eye_line[1], eye_line[0])
+    tilt_deg = 180 * tilt / np.pi
+
+    scale = OUT_EYE_SPACE / abs(eye0[0] - eye1[0])
+    pimgs = pimg.resize((int(iw * scale), int(ih * scale)), resample=PImage.Resampling.LANCZOS)
+
+    # rotate around nose
+    new_mid = [int(c * scale) for c in mid]
+    crop_box = (new_mid[0] - (OUT_W // 2),
+                new_mid[1] - OUT_NOSE_TOP,
+                new_mid[0] + (OUT_W // 2),
+                new_mid[1] + (OUT_H - OUT_NOSE_TOP))
+
+    img_out = pimgs.rotate(tilt_deg, center=new_mid, resample=PImage.Resampling.BICUBIC).crop(crop_box)
+    out_images.append(gr.Image(img_out, visible=True))
+
+  out_images += out_pad
+  return out_images[:NUM_OUTS]
+
+
+with gr.Blocks() as demo:
+  gr.Markdown("""
+              # 9103H 2024F Face Alignment Tool.
+              ## Interface for face detection, alignment, cropping\
+              to help create dataset for [HWXX](https://github.com/DM-GY-9103-2024F-H/).
+              """)
+
+  gr.Interface(
+    face,
+    inputs=gr.Image(type="pil"),
+    outputs=all_outputs,
+    cache_examples=True,
+    examples=[["./imgs/03.webp"], ["./imgs/11.jpg"]]
+  )
+
+if __name__ == "__main__":
+   demo.launch()

models/haarcascade_frontalface_alt2.xml

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b0c967d9abbdfbde025eb9c786947d151b6426040d07a8f9562ed8fd90724b4
+size 540616

models/lbfmodel.yaml

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70dd8b1657c42d1595d6bd13d97d932877b3bed54a95d3c4733a0f740d1fd66b
+size 56375857

requirements.txt

@@ -0,0 +1,2 @@
+opencv-contrib-python
+opencv-python