Merge branch 'faces'

.gitignore

@@ -179,3 +179,5 @@ $RECYCLE.BIN/
 # End of https://www.toptal.com/developers/gitignore/api/windows,python
 
 !/src/lib/
+autogen_buildtime.py
+*.png

Dockerfile

@@ -1,11 +1,14 @@
 # Based off of https://github.com/minimaxir/gpt-2-cloud-run/blob/master/Dockerfile
-FROM python:3.8-slim-buster
+FROM python:3.10-slim-buster
 
 WORKDIR /code
 
-RUN pip3 --no-cache-dir install -U py-cord==2.0.0b5 python-dotenv
+# opencv dependencies https://stackoverflow.com/a/63377623
+RUN apt-get update && apt-get install -y ffmpeg libsm6 libxext6 && apt-get clean && rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
 
-COPY .env src/ secret-scholars-bot-config.json ./
+RUN pip3 --no-cache-dir install -U py-cord python-dotenv markovchain pillow opencv-python
+
+COPY .env src/ secret-scholars-bot-config.json markov.txt ./
 
 RUN printf "build_time=%s" "'`date -Iseconds`'" > autogen_buildtime.py
 

Dockerfile-arm32v6

@@ -1,14 +1,16 @@
 # Based off of https://github.com/minimaxir/gpt-2-cloud-run/blob/master/Dockerfile
 FROM arm32v6/python:3.10-alpine
 
-RUN apk update && apk add gcc musl-dev
+RUN apk update && apk add gcc musl-dev py3-opencv zlib-dev jpeg-dev
 
 WORKDIR /code
 
-RUN pip3 --no-cache-dir install -U py-cord python-dotenv markovchain
+ENV PYTHONPATH=$PYTHONPATH:/usr/lib/python3.10/site-packages
+
+RUN pip3 --no-cache-dir install -U py-cord python-dotenv markovchain pillow
 
 COPY .env src/ secret-scholars-bot-config.json markov.txt ./
 
 RUN printf "build_time=%s" "'`date -Iseconds`'" > autogen_buildtime.py
 
-CMD [ "python", "./bot.py" ]
+# CMD [ "python", "./bot.py" ]

README.md

@@ -35,5 +35,5 @@ docker run -d secret-scholars-bot
 ### ARM32v6
 Run
 ```sh
-docker run -d secret-scholars-bot secret-scholars-bot:arm32v6
+docker run -d secret-scholars-bot:arm32v6
 ```

requirements.txt

@@ -0,0 +1,6 @@
+opencv-python
+numpy
+pillow
+pycord
+python-dotenv
+discord

src/bot.py

@@ -5,6 +5,7 @@
 from datetime import datetime, timezone
 import donut
 import soup
+import faces
 import roll
 import config
 import asyncio
@@ -49,6 +50,7 @@ async def on_message(message):
     await soup.on_message(message, client, config)
     await excuse.on_message(message)
     await markov.on_message(message)
+    await faces.on_message(message, client, config)
     if await roll.on_message(message):
       return
 

src/config.py.example

@@ -20,3 +20,12 @@ excuses = [
   ("grocery shopping", 2)
 ] # An array of string messages or tuples containing a string message and a weight. Non-tuples have a weight on 1.
 # Weights can be a decimal, 0 < w < infinity
+
+# Face configs
+face_picture = 'lib/face.png'
+
+face_dims = [
+    [43, 208, 430, 430],
+    [245, 125, 82, 82],
+    [98, 131, 87, 87]
+]

src/faces.py

@@ -0,0 +1,31 @@
+import lib.faces_util as faces_util
+import os
+import discord
+
+BYTES_IN_MEGABYTE = 1048576
+
+allowable_attachment_types = ['image/png', 'image/jpeg']
+
+async def on_message(message, client, conf):
+  if message.content.startswith("$face"):
+    if len(message.attachments) > 0:
+      for attachment in message.attachments:
+        # check if attachment is a picture and is smaller than 8mb
+        if attachment.content_type in allowable_attachment_types and attachment.size < (8 * BYTES_IN_MEGABYTE):
+          if attachment.proxy_url != '404':
+            try:
+              await attachment.save('attachment_picture.png', use_cached=True)
+              found, path = faces_util.get_face_replace('attachment_picture.png')
+              if found:
+                await message.channel.send(file=discord.File(path))
+              else:
+                await message.channel.send('No face found :(')
+              if os.path.exists('face_detected.png'): 
+                os.remove('face_detected.png')
+            except discord.HTTPException as error:
+              logging.exception(f'{datetime.now(timezone.utc)} Face on_message')
+              await message.channel.send('Could not download attachment :(')
+            except discord.NotFound as error:
+              logging.exception(f'{datetime.now(timezone.utc)} Face on_message')
+              await message.channel.send('Attachment not found :(')
+

src/lib/faces_util.py

@@ -0,0 +1,130 @@
+import cv2
+import numpy as np
+import math
+import aiohttp
+import os
+import config as conf
+from PIL import Image
+
+face_cascade = cv2.CascadeClassifier('lib/opencv/face_detector.xml')
+eye_cascade = cv2.CascadeClassifier('lib/opencv/eye_detector.xml')
+smile_cascade = cv2.CascadeClassifier('lib/opencv/smile_detector.xml')
+
+def load_image_rgba(path):
+  og_image = Image.open(path)
+  rgba_image = og_image.convert(mode='RGBA')
+  img = np.array(rgba_image)
+  # B and R channels are swapped and there didn't seem to be a better way to do this
+  copy = np.array(img)
+  img[:,:,0] = copy[:,:,2]
+  img[:,:,2] = copy[:,:,0]
+  return img
+
+# https://www.geeksforgeeks.org/python-smile-detection-using-opencv/
+def box_faces(gray, frame):
+  faces = face_cascade.detectMultiScale(gray, 1.3, 5)
+  for (x, y, w, h) in faces:
+    # print(f'{x}, {y}, {w}, {h}')
+    cv2.rectangle(frame, (x, y), ((x + w), (y + h)), (255, 0, 0, 255), 2)
+    roi_gray = gray[y:y + h, x:x + w]
+    roi_color = frame[y:y + h, x:x + w]
+    # smiles = smile_cascade.detectMultiScale(roi_gray, 1.8, 20)
+    eyes = eye_cascade.detectMultiScale(roi_gray)
+
+    for (sx, sy, sw, sh) in eyes:
+      # print(f'{sx}, {sy}, {sw}, {sh}')
+      cv2.rectangle(roi_color, (sx, sy), ((sx + sw), (sy + sh)), (0, 0, 255, 255), 2)
+  return frame
+
+# https://stackoverflow.com/a/59211216
+def rgba_overlay(background, foreground):
+  alpha_background = background[:,:,3] / 255.0
+  alpha_foreground = foreground[:,:,3] / 255.0
+
+  # set adjusted colors
+  for color in range(0, 3):
+      background[:,:,color] = alpha_foreground * foreground[:,:,color] + \
+          alpha_background * background[:,:,color] * (1 - alpha_foreground)
+
+  # set adjusted alpha and denormalize back to 0-255
+  background[:,:,3] = (1 - (1 - alpha_foreground) * (1 - alpha_background)) * 255
+
+def replace_faces(gray, frame, replacement, replacement_dims):
+  faces = face_cascade.detectMultiScale(gray, 1.1, 5)
+  base_w = frame.shape[1]
+  base_h = frame.shape[0]
+  found = False
+  for (x, y, w, h) in faces:
+    found = True
+    overlay = np.zeros(frame.shape)
+    replacement_scale_x = w / replacement_dims[0][2]
+    replacement_scale_y = h / replacement_dims[0][3]
+    scaled_replacement_x = x - math.ceil(replacement_dims[0][0] * replacement_scale_x)
+    scaled_replacement_y = y - math.ceil(replacement_dims[0][1] * replacement_scale_y)
+    scaled_replacement_x = scaled_replacement_x if scaled_replacement_x > 0 else 0
+    scaled_replacement_y = scaled_replacement_y if scaled_replacement_y > 0 else 0
+    scaled_full_w = math.ceil(replacement.shape[1] * replacement_scale_x)
+    scaled_full_h = math.ceil(replacement.shape[0] * replacement_scale_y)
+    resized = cv2.resize(replacement, (scaled_full_w, scaled_full_h), interpolation=cv2.INTER_CUBIC)
+    up = base_h - (scaled_full_h + scaled_replacement_y)
+    up = -scaled_replacement_y if scaled_replacement_y + up < 0 else 0
+    left = base_w - (scaled_full_w + scaled_replacement_x)
+    left = -scaled_replacement_x if scaled_replacement_x + left < 0 else 0
+    insert_w = base_w - scaled_replacement_x if scaled_replacement_x + scaled_full_w > base_w else scaled_full_w
+    insert_h = base_h - scaled_replacement_y - up if scaled_replacement_y + scaled_full_h + up > base_h else scaled_full_h
+    overlay[scaled_replacement_y + up:scaled_replacement_y + insert_h + up, scaled_replacement_x + left:scaled_replacement_x + insert_w + left] = resized[0:insert_h, 0:insert_w]
+    rgba_overlay(frame, overlay)
+
+  return found, frame
+
+def get_face_replace(replacing_image):
+  face_replace = load_image_rgba(conf.face_picture)
+  attachment_pic = load_image_rgba(replacing_image)
+  gray = cv2.cvtColor(attachment_pic, cv2.COLOR_RGBA2GRAY)
+  found,img = replace_faces(gray, attachment_pic, face_replace, conf.face_dims)
+
+  if found:
+    cv2.imwrite('face_detected.png', img)
+    return True, 'face_detected.png'
+
+  return False, None
+
+if __name__ == '__main__':
+  img = load_image_rgba('student_small.png')
+  # img = load_image_rgba('group.png')
+  student = np.array(img)
+
+  face = load_image_rgba('face.png')
+  face_dims = [
+    [43, 208, 430, 430],
+    [245, 125, 82, 82],
+    [98, 131, 87, 87]
+  ]
+  gray = cv2.cvtColor(img, cv2.COLOR_RGBA2GRAY)
+  found,img = replace_faces(gray, img, face, face_dims)
+  cv2.imwrite("face_detected.png", img)
+  print('Successfully saved')
+
+# A video with faces applied for testing purposes
+# video_capture = cv2.VideoCapture(0)
+# while video_capture.isOpened():
+# # Captures video_capture frame by frame
+#   _, frame = video_capture.read()
+
+#   # To capture image in monochrome				
+#   gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+
+#   # calls the box_faces() function
+#   canvas = box_faces(gray, frame)
+
+#   # Displays the result on camera feed					
+#   cv2.imshow('Video', canvas)
+
+#   # The control breaks once q key is pressed					
+#   if cv2.waitKey(1) & 0xff == ord('q'):			
+#     break
+
+# # Release the capture once all the processing is done.
+# video_capture.release()								
+# cv2.destroyAllWindows()
+