Initial sheets bill commit

.gitignore

@@ -6,6 +6,8 @@ data/*
 !data/sample_images
 !data/sample_test
 
+reports/
+
 __pycache__
 
 # Under work

counter.py

@@ -0,0 +1,42 @@
+import os
+
+from sympy import Id
+
+class IDCounter(object):
+  def __init__(self, threshold, ID, sessionId, path):
+    self.count_thresh = threshold
+    self.names = [f"{entity['first']} {entity['last']}" for entity in ID]
+    self.counters = {i:0 for i in self.names}
+    self.attendance = {i:False for i in self.names}
+
+    self.sessionId = sessionId
+    self.path = path
+    with open(path, 'wb') as f:
+      f.write(f"Attendance report for {sessionId}\n\nRegistered:\n".encode("utf-8"))
+
+  def write(self, val):
+    with open(self.path, 'ab') as f:
+      f.write(val.encode("utf-8"))
+
+  def update(self, id):
+    if id in self.counters:
+      self.counters[id] += 1
+      if self.counters[id] >= self.count_thresh:
+        if self.attendance[id] == False:
+          self.attendance[id] = True
+          self.write(f"✔\t{id}\n")
+        return True
+    else:
+      self.counters[id] = 1
+    return False
+
+  def showReport(self):
+    append = f"\nNot registered:\n"
+    for person in self.attendance:
+      if self.attendance[person] == False:
+        append += f"❌\t{person}\n"
+    append += '\n\n[TECLARS ATTENDANCE REPORT COMPLETE]\n'
+
+    self.write(append)
+
+    os.system(f"notepad.exe {self.path}")

get_embed.py → embed.py

@@ -1,14 +1,30 @@
 from facenet_pytorch import MTCNN, InceptionResnetV1
 from tqdm import tqdm
 from PIL import Image
+import argparse
 import torch
 import os
 import json
 
-with open('data\id.json', 'r') as f:
+parser = argparse.ArgumentParser(description='TECLARS Embedding Generator')
+parser.add_argument('-i', '--id_path', type=str, default='./data/id.json',
+                    help='Path of ID data JSON file')
+parser.add_argument('-p', '--img_dir', type=str, default='./data/images',
+                    help='Directory of where image data is stored')
+parser.add_argument('-o', '--out_path', type=str, default='./data/embeddings.pt',
+                    help='Path of ID data JSON file')
+parser.add_argument('-d', '--device', type=str, default='auto',
+                    help='Device to compute algorithm on')
+
+args = parser.parse_args()
+
+with open(args.id_path, 'r') as f:
     ID = json.load(f)
 
-device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
+if args.device == "auto":
+    device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
+else:
+    device = torch.device(args.device)
 print('Running on device: {}'.format(device))
 
 print("Aligning for faces from data")
@@ -17,7 +33,7 @@
 
 aligned = []
 for idx, person in tqdm(enumerate(ID)):
-    fpath = os.path.join('./data/images', person['image'])
+    fpath = os.path.join(args.img_dir, person['image'])
 
     with Image.open(fpath) as x:
         x_aligned, prob = mtcnn(x, return_prob=True)
@@ -34,6 +50,6 @@
 print("Computing face embeddings")
 aligned = torch.stack(aligned).to(device)
 embeddings = resnet(aligned).detach().cpu()
-torch.save({'embedding': embeddings}, 'data/embeddings.pt')
+torch.save({'embedding': embeddings}, args.out_path)
 
 print("\n[All done]")

run.py

@@ -1,3 +1,4 @@
+from genericpath import exists
 from facenet_pytorch import MTCNN, InceptionResnetV1
 import torch
 import torch.nn as nn
@@ -9,6 +10,9 @@
 import time
 import argparse
 
+from counter import IDCounter
+
+
 def euclidean_distance(out, refs):
     return (out - refs).norm(dim=1)
 
@@ -75,6 +79,15 @@ def identify_faces(pil_image):
     return None
 
 def video():
+    if args.session_id is None:
+        sess_id = "sheets-bill-" + input("Enter session ID: sheets-bill-")
+    else:
+        sess_id = "sheets-bill-" + args.session_id
+
+    os.makedirs(args.output_dir, exist_ok=True)
+    report_path = os.path.join(args.output_dir, f"{sess_id}.txt")
+    counter = IDCounter(args.number, ID, sess_id, report_path)
+
     print("Starting video capture\n")
     video_capture = cv2.VideoCapture(args.camera, cv2.CAP_DSHOW)
 
@@ -103,13 +116,17 @@ def video():
                 if args.show_unrecognised and entity is None:
                     cv2.rectangle(frame, (bounds[0], bounds[1]), (bounds[2], bounds[3]), (0, 0, 255), 2)
                 else:
-                    text = f"{entity['first']} {entity['last']}: {round(confidence*100, 2)}%"
-                    cv2.putText(frame, text, (bounds[0], bounds[1]-10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)
-                    cv2.rectangle(frame, (bounds[0], bounds[1]), (bounds[2], bounds[3]), (0, 255, 0), 2)
-
-                    if entity not in registered_students:
-                        print(f"{entity['first']} {entity['last']} registered")
-                        registered_students.append(entity)
+                    name = f"{entity['first']} {entity['last']}"
+                    if counter.update(name):
+                        if entity not in registered_students:
+                            print(f"{entity['first']} {entity['last']} registered")
+                            registered_students.append(entity)
+                        color = (0, 255, 0)
+                    else:
+                        color = (0, 210, 255)
+                    text = f"{name}: {round(confidence*100, 2)}%"
+                    cv2.putText(frame, text, (bounds[0], bounds[1]-10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2)
+                    cv2.rectangle(frame, (bounds[0], bounds[1]), (bounds[2], bounds[3]), color, 2)
 
         cv2.imshow('TECLARS Main UI', frame)
 
@@ -128,6 +145,8 @@ def video():
         for e, entity in enumerate(registered_students)
     ]))
 
+    counter.showReport()
+
 def test():
     print("Beginning testing")
     for filename in os.listdir(args.test_path):
@@ -188,6 +207,8 @@ def check():
 
 parser.add_argument('-r', '--threshold', type=float, default=0.8,
                     help='Probability above which a face will be considered recognised')
+parser.add_argument('-n', '--number', type=int, default=5,
+                    help='Minimum number of frames above which a face will be considered recognised')
 parser.add_argument('-g', '--margin', type=float, default=0.1,
                     help='Minimum probability margin above next likely face for the face to be considered recognised')
 parser.add_argument('-t', '--temp', type=float, default=2,
@@ -198,10 +219,14 @@ def check():
                     help='Device to compute algorithm on')
 parser.add_argument('-m', '--mode', type=str, choices=['cosine', 'euclidean'], default='cosine',
                     help='Distance function for evaluating the similarity between face embeddings')
-parser.add_argument('-u', '--show_unrecognised', action="store_true",
+parser.add_argument('-u', '--show_unrecognised', action="store_false",
                     help='Remove bounding boxes around unrecognised faces')
 parser.add_argument('-i', '--ignore', type=int, default=100,
                     help='Ignore faraway faces with a width smaller than this value (set 0 to include all faces)')
+parser.add_argument('-s', '--session_id', type=str, default=None,
+                    help="Session ID")
+parser.add_argument('-o', '--output_dir', type=str, default="./reports",
+                    help="Directory to output reports")
 parser.add_argument('-x', '--dev', action='store_true',
                     help="Enable developer options")