chandrikadeb7 / face-mask-detection

Face Mask Detection system based on computer vision and deep learning using OpenCV and Tensorflow/Keras

Hi I'm Chandrika Deb

git commit -m "Hello, I'm a Software Developer at Amdocs"

Liked my Contributions:question::point_right: Nominate Me for GitHub Stars ⭐ ✨

Liked my GitHub:question::point_right: Sponsor me a Dollar 💰 💵 ✨

• 👨‍💻 All about me is at My Website

• 📝 Sometimes I write articles on Medium

• 💬 Ask me about var topics = [ "Java", "Spring", "Python", "CSS" ]

• 📂 Purchase Face Mask Detection Report and Slides on Gumroad

• 📫 How to reach me [email protected]

• 📄 Know about my work & experiences My Resume

What are my featured projects:question::rocket:

100DaysOfCode⌛
Face Mask Detection😷
GirlScript Twitter Bot🤖

Wanna see my blogs:question::fire:

• Connect your Chatwoot to Netlify integration in 10 minutes
• How my 22 month developer experience at Amdocs shaped my career?
• Connect your Chatwoot to Netlify integration in 10 minutes
• SAP Testing Best Practices
• Using MDX with Docz Has Never Been This Easy!

What can I help with:question::cyclone:

git commit -m "All about GitHub!" 😁

Connect with me:

Languages and Tools:

Support:

face-mask-detection's Issues

Fix accuracy of the model

The model gives false results sometimes for "with mask" when face is covered with hands.

Add Contributors.md file

Hai,
I would like to add Contributors.md file in your project in the form of a table and display a link to it in README.md

Update number of images in Dataset section of Readme

Update the dataset section with present number of images in dataset folder of the repo.

Integrate welcome bot

I can add a welcome bot config file having a proper message that will show up when any user will open up an issue or pull-request for the first time.
For reference, check out: https://github.com/apps/welcome

If you feel this is a great idea!
Please assign it to me.

Alarm system when "No Mask" detected

Implement an alarming system when the camera detects "No Mask" in real-time video streams.

Video stream from Ip camera

Please add the ability to process a video from an IP camera...

RTSP protocol will be great...

model for street view

Hi @chandrikadeb7 , i need dataset or model support street view detection mask or drone view . any help you can provide would be greatly appreciated. thanks

Not Detecting Multiple Faces in Video

When i tried to detect multiple faces while running the video version, it was only able to detect one face at a time.
And when i ran the image version it was successfully able to detect multiple faces and tell whether they were wearing mask or not.
Can you tell me why this is happening? Because its suppose to detect multiple faces. And what could be it's solution.
It would be really great if you could help me on this.
Thanks

Display ratio / count of who have wore the mask out of persons recognised

Displaying the total count/ratio of persons who have wore mask out of total count of persons recognised .

Face with Beard

In case of beard algorithm predict face with Mask but there is no mask on face

Create Code of Conduct. md

Hi,
I found that this project repository is not having a code of conduct file in this repository, kindly assign me this issue and I would like to contribute a well-formatted code of conduct file as you know code of conduct is a necessary file which the contributors need to follow so that everyone follows a decorum.

Help with Working Instructions

When I run working commands, I got the following errors:

C:\Users\Ahmed Ahmed\Face-Mask-Detection>python train_mask_detector.py --dataset dataset
2020-10-30 21:33:55.095518: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cudart64_101.dll
Traceback (most recent call last):
File "train_mask_detector.py", line 5, in
from tensorflow.keras.preprocessing.image import ImageDataGenerator
File "C:\Program Files\Python38\lib\site-packages\tensorflow_init_.py", line 49, in
from ._api.v2 import operators
ImportError: cannot import name 'operators' from 'tensorflow.api.v2' (C:\Program Files\Python38\lib\site-packages\tensorflow_api\v2_init.py)

C:\Users\Ahmed Ahmed\Face-Mask-Detection>python detect_mask_video.py
2020-10-30 21:34:38.028712: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cudart64_101.dll
Traceback (most recent call last):
File "detect_mask_video.py", line 5, in
from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
File "C:\Program Files\Python38\lib\site-packages\tensorflow_init_.py", line 49, in
from ._api.v2 import operators
ImportError: cannot import name 'operators' from 'tensorflow.api.v2' (C:\Program Files\Python38\lib\site-packages\tensorflow_api\v2_init.py)

Implementation in Rasp Pi

Please create a go-to-code folder comprising the necessary codes to detect masks using Rasp Pi.

chenge size

hello i want to change out put size in video detection in this line
frame = imutils.resize(frame, width=400)
can change width but i cant find height

Collect more dataset samples of real with_mask images

The images must not be duplicate of the current dataset

[SCI] Benchmarking different architectures

For object detection, it would be interesting for participants to consider different models, e.g. YOLOv4 (https://arxiv.org/abs/2004.10934), YOLOv5 (https://github.com/ultralytics/yolov5), EfficientDet (https://arxiv.org/abs/1911.09070). Implementations of these models in Tensorflow, PyTorch and Darkent are readily available.

Moving the Arduino motor to open door if the Mask is detected

I want to put hands on to open door at entrance if a person is wearing mask

0 Accuracy in CCTV footage

As the code is intended to identify the mask from any video stream, but the major issue is that it is not able to identify "FACE MASK" in the live feeds from the CCTV / DVR / NVR.

Web application for the project

Please make a PR for Web application of the project.

Not all images are able to process

if I process the below image it gives an error of Assertion failed in OpenCV.

Error:
File "detect_mask_image.py", line 105, in
mask_image()
File "detect_mask_image.py", line 76, in mask_image
face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)
cv2.error: OpenCV(4.2.0) C:\projects\opencv-python\opencv\modules\imgproc\src\color.cpp:182: error: (-215:Assertion failed) !_src.empty() in function 'cv::cvtColor'

If you encounter this error you need to manually delete "./images/out.jpg" and save another image which can be processed e.g. "./images/pic2.jpg" as "./images/out.jpg".

Add new entries for coloured masks in dataset folder

ffmpeg library should be installed in order to run detect_mask_video.py

Make the complete Wiki of the repo describing the project pipeline and architecture.

accuracy

Hi all,
I tried to test this model with different sizes of dataset. What I've done is to randomly select 1, 5, 10 or 20 % of the original dataset and training the model with train_mask_detector.py, but the accuracy value that is shown in the end of training is very odd. these are my results after training the model three times ( because I thought these numbers are wrong and wanted to make sure it gave me same numbers consistent before asking here ) :

first time :
10 % :
0.974025974025974
20 % :
0.974025974025974

---

second time :
10 % :
0.987012987012987
20 % :
0.9675324675324676

---

third time :
10 % :
0.974025974025974
20 % :
0.961038961038961

and one time when I was training with 5% it gave me the accuracy of 1.0
also called the sklearn.metrics.acuuracy_score which is the same function that is called after training under "[INFO] evaluating network..."
can anyone help me understand what I'm doing wrong or something??
thank you all

Improve the documentation by adding awards/achievements section

Please share a defined layout for the Readme documentation of this repo. For the details to be added, I'll send the necessary images.

mask detection method

Hi, what type of classifier are you using for mask detection?

typos identified and fixed and huge difference in accuracy on google colab

some errors faced and resolved

file: train_mask_detector.py

line no. 149 plt.plot(np.arange(0, N), H.history["acc"], label="train_acc")
it gives KeyError.

solution:
plt.plot(np.arange(0, N), H.history["accuracy"], label="train_acc")

similarly
line no. 150 plt.plot(np.arange(0, N), H.history["val_acc"], label="val_acc")
it also gives KeyError.

solution:
plt.plot(np.arange(0, N), H.history["val_accuracy"], label="val_acc")

Your results state that your model achieved 93% accuracy but when running on google colab, tensorflow==2.3.0, I have achieved 99% accuracy

video stops frequently with error (-215:Assertion failed) !_src.empty() in function 'cvtColor'

How can I use IP camera for streaming and video stops frequently with following error.

Traceback (most recent call last):
File "detect_mask_video.py", line 116, in
(locs, preds) = detect_and_predict_mask(frame, faceNet, maskNet)
File "detect_mask_video.py", line 55, in detect_and_predict_mask
face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)
cv2.error: OpenCV(4.2.0) /Users/travis/build/skvark/opencv-python/opencv/modules/imgproc/src/color.cpp:182: error: (-215:Assertion failed) !_src.empty() in function 'cvtColor'

same code

I found same code but different github.
https://github.com/micaelillos/Mask-Detection
who is first?

Create a help guide for running the project in Windows OS

Documentation should include detailed steps including screenshots.

Face Mask

cpu speed

Hi
I have a quesion about inference speed. I check inference speed in cpu
Then, speed is 30~40ms. I think it is too slow while using mobilenetv2.
Is this speed correct?

when the object is too large in the picture,program will crash

try to add the borders of picture to can be a solution.

Using inception v3 architecture to train the model{SWOC21}

Inception V3 can provide better accuracies to the model.So can you assign me this issue?

Google collab

Can we do the same project in Google collab ? Implementing this in google collab will help many students to learn this project .

difficulty detecting "black" hex:#000000 colored masks on raspberry pi pycamera

permission required

while executing this command python3 train_mask_detector.py --dataset dataset its showing permission denied so if possible rectify this issue

multiple faces

hi, multiple faces in the real-time video not detected?

EdgeTPU compatible model (Google Coral)

Hey there, good work ! Do you have the model that is compatible with the Coral?

Face Mask detection Error

Getting following error
sudo python3 detect_mask_video.py
2020-11-17 14:44:36.808007: E tensorflow/core/platform/hadoop/hadoop_file_system.cc:132] HadoopFileSystem load error: libhdfs.so: cannot open shared object file: No such file or directory
..............
Traceback (most recent call last):
File "detect_mask_video.py", line 98, in
maskNet = load_model(args["model"])
File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/keras/saving/save.py", line 149, in load_model
loader_impl.parse_saved_model(filepath)
File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/saved_model/loader_impl.py", line 83, in parse_saved_model
constants.SAVED_MODEL_FILENAME_PB))
OSError: SavedModel file does not exist at: mask_detector.model/{saved_model.pbtxt|saved_model.pb}

Also getting following error
(fmdt1-env) pi@raspberrypi:/FMDet1 $ sudo pip3 install -r requirements.txt
Looking in indexes: https://pypi.org/simple, https://www.piwheels.org/simple
ERROR: Could not find a version that satisfies the requirement tensorflow>=1.15.2 (from -r requirements.txt (line 1)) (from versions: 0.11.0, 1.12.0, 1.13.1, 1.14.0)
ERROR: No matching distribution found for tensorflow>=1.15.2 (from -r requirements.txt (line 1))
(fmdt1-env) pi@raspberrypi:/FMDet1 $ sudo pip3 show tensorflow

Accuracy - Mask and Withoutmask %

There is one issue with this model or prediction.

The % of the mask and withoutmask % is inconsistent. It varies to the same position pf the mask.

What is the solution for this ?...

Collect images of coloured masked faces

Please collect more images of black or coloured masked faces to be included in the dataset.

Use case

Hello, I am a beginner web developer and I would love to create a project similar to yours.

However, because I'm a beginner, I don't know if the MIT license allows me to copy your software or if I need additional permission or something else.

If I can make a copy of your source code, do I need to include thanks, pay for something, or anything else?

Thank you :)

Creating an automated mask dispenser

Implementation of an automated mask dispenser that dispenses mask for any person who is not wearing a mask. This kind of solution can be used by shopping complexes etc.

Deployment Issues

Hi!

I have a simple question. How did you manage to deploy this on Heroku? I tried doing the same but faced a slug size issue.

Multiple Face detection

The code is working perfectly but there is one major issues that are the system is not able to detect multiple faces.
As the name suggests it should be able to detect multiple faces.

KeyError: 'acc' when saving model in Google Colab

I was using Google Colab to train a model so I could try it on my laptop, however, when saving the model this error occurs, I'm using a GPU

[INFO] evaluating network...
              precision    recall  f1-score   support

   with_mask       0.99      1.00      0.99       384
without_mask       1.00      0.99      0.99       386

    accuracy                           0.99       770
   macro avg       0.99      0.99      0.99       770
weighted avg       0.99      0.99      0.99       770

[INFO] saving mask detector model...
Traceback (most recent call last):
  File "train_mask_detector.py", line 149, in <module>
    plt.plot(np.arange(0, N), H.history["acc"], label="train_acc")
KeyError: 'acc'

Previous Steps

• Cloned the repository

• Downloaded requirements

• Ran python3 train_mask_detector.py --dataset dataset

Verbose Output

2020-08-15 15:09:08.012494: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.10.1
[INFO] loading images...
/usr/local/lib/python3.6/dist-packages/PIL/Image.py:961: UserWarning: Palette images with Transparency expressed in bytes should be converted to RGBA images
  "Palette images with Transparency expressed in bytes should be "
tcmalloc: large alloc 2315722752 bytes == 0x91b80000 @  0x7f2a8385e1e7 0x7f2a8007a5e1 0x7f2a800dee88 0x7f2a800e1fc8 0x7f2a800e25a5 0x7f2a8017984d 0x50a7f5 0x50cfd6 0x507f24 0x50b053 0x634dd2 0x634e87 0x63863f 0x6391e1 0x4b0dc0 0x7f2a8345bb97 0x5b26fa
tcmalloc: large alloc 1852096512 bytes == 0x11bbf2000 @  0x7f2a8385e1e7 0x7f2a8007a5e1 0x7f2a800dee88 0x7f2a800defa3 0x7f2a8016a0c6 0x7f2a8016a528 0x50c4de 0x507f24 0x509c50 0x50a64d 0x50cfd6 0x507f24 0x509c50 0x50a64d 0x50c1f4 0x58e809 0x4c9306 0x588114 0x58834e 0x5517c1 0x5a9eec 0x50a783 0x50c1f4 0x507f24 0x509c50 0x50a64d 0x50cfd6 0x507f24 0x50b053 0x634dd2 0x634e87
WARNING:tensorflow:`input_shape` is undefined or non-square, or `rows` is not in [96, 128, 160, 192, 224]. Weights for input shape (224, 224) will be loaded as the default.
2020-08-15 15:09:31.941657: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcuda.so.1
2020-08-15 15:09:32.004988: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:982] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-08-15 15:09:32.005892: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1716] Found device 0 with properties: 
pciBusID: 0000:00:04.0 name: Tesla K80 computeCapability: 3.7
coreClock: 0.8235GHz coreCount: 13 deviceMemorySize: 11.17GiB deviceMemoryBandwidth: 223.96GiB/s
2020-08-15 15:09:32.005987: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.10.1
2020-08-15 15:09:32.260469: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcublas.so.10
2020-08-15 15:09:32.402794: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcufft.so.10
2020-08-15 15:09:32.430120: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcurand.so.10
2020-08-15 15:09:32.701520: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcusolver.so.10
2020-08-15 15:09:32.728638: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcusparse.so.10
2020-08-15 15:09:33.266461: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudnn.so.7
2020-08-15 15:09:33.266776: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:982] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-08-15 15:09:33.267816: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:982] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-08-15 15:09:33.268607: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1858] Adding visible gpu devices: 0
2020-08-15 15:09:33.286410: I tensorflow/core/platform/profile_utils/cpu_utils.cc:104] CPU Frequency: 2300000000 Hz
2020-08-15 15:09:33.286728: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x1f08d80 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-08-15 15:09:33.286778: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version
2020-08-15 15:09:33.373932: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:982] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-08-15 15:09:33.374835: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x1f08bc0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2020-08-15 15:09:33.374876: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Tesla K80, Compute Capability 3.7
2020-08-15 15:09:33.376571: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:982] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-08-15 15:09:33.377366: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1716] Found device 0 with properties: 
pciBusID: 0000:00:04.0 name: Tesla K80 computeCapability: 3.7
coreClock: 0.8235GHz coreCount: 13 deviceMemorySize: 11.17GiB deviceMemoryBandwidth: 223.96GiB/s
2020-08-15 15:09:33.377428: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.10.1
2020-08-15 15:09:33.377501: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcublas.so.10
2020-08-15 15:09:33.377562: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcufft.so.10
2020-08-15 15:09:33.377609: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcurand.so.10
2020-08-15 15:09:33.377653: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcusolver.so.10
2020-08-15 15:09:33.377702: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcusparse.so.10
2020-08-15 15:09:33.377751: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudnn.so.7
2020-08-15 15:09:33.377909: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:982] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-08-15 15:09:33.378782: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:982] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-08-15 15:09:33.379553: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1858] Adding visible gpu devices: 0
2020-08-15 15:09:33.383743: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.10.1
2020-08-15 15:09:37.202833: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1257] Device interconnect StreamExecutor with strength 1 edge matrix:
2020-08-15 15:09:37.202895: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1263]      0 
2020-08-15 15:09:37.202914: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1276] 0:   N 
2020-08-15 15:09:37.207175: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:982] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-08-15 15:09:37.208107: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:982] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-08-15 15:09:37.208872: W tensorflow/core/common_runtime/gpu/gpu_bfc_allocator.cc:39] Overriding allow_growth setting because the TF_FORCE_GPU_ALLOW_GROWTH environment variable is set. Original config value was 0.
2020-08-15 15:09:37.208930: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1402] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10630 MB memory) -> physical GPU (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0, compute capability: 3.7)
Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet_v2/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_224_no_top.h5
9412608/9406464 [==============================] - 0s 0us/step
[INFO] compiling model...
[INFO] training head...
Epoch 1/20
2020-08-15 15:09:42.573008: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcublas.so.10
2020-08-15 15:09:44.011634: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudnn.so.7
96/96 [==============================] - 39s 408ms/step - loss: 0.3189 - accuracy: 0.8561 - val_loss: 0.0953 - val_accuracy: 0.9766
Epoch 2/20
96/96 [==============================] - 37s 390ms/step - loss: 0.1159 - accuracy: 0.9599 - val_loss: 0.0671 - val_accuracy: 0.9805
Epoch 3/20
96/96 [==============================] - 37s 389ms/step - loss: 0.0820 - accuracy: 0.9708 - val_loss: 0.0646 - val_accuracy: 0.9779
Epoch 4/20
96/96 [==============================] - 37s 381ms/step - loss: 0.0666 - accuracy: 0.9773 - val_loss: 0.0505 - val_accuracy: 0.9857
Epoch 5/20
96/96 [==============================] - 36s 378ms/step - loss: 0.0577 - accuracy: 0.9816 - val_loss: 0.0436 - val_accuracy: 0.9870
Epoch 6/20
96/96 [==============================] - 37s 388ms/step - loss: 0.0547 - accuracy: 0.9793 - val_loss: 0.0441 - val_accuracy: 0.9883
Epoch 7/20
96/96 [==============================] - 37s 385ms/step - loss: 0.0489 - accuracy: 0.9826 - val_loss: 0.0432 - val_accuracy: 0.9883
Epoch 8/20
96/96 [==============================] - 37s 381ms/step - loss: 0.0472 - accuracy: 0.9855 - val_loss: 0.0407 - val_accuracy: 0.9896
Epoch 9/20
96/96 [==============================] - 37s 380ms/step - loss: 0.0443 - accuracy: 0.9846 - val_loss: 0.0364 - val_accuracy: 0.9909
Epoch 10/20
96/96 [==============================] - 37s 388ms/step - loss: 0.0362 - accuracy: 0.9872 - val_loss: 0.0357 - val_accuracy: 0.9896
Epoch 11/20
96/96 [==============================] - 38s 391ms/step - loss: 0.0393 - accuracy: 0.9878 - val_loss: 0.0337 - val_accuracy: 0.9909
Epoch 12/20
96/96 [==============================] - 36s 378ms/step - loss: 0.0342 - accuracy: 0.9898 - val_loss: 0.0375 - val_accuracy: 0.9883
Epoch 13/20
96/96 [==============================] - 36s 378ms/step - loss: 0.0324 - accuracy: 0.9895 - val_loss: 0.0329 - val_accuracy: 0.9896
Epoch 14/20
96/96 [==============================] - 37s 386ms/step - loss: 0.0288 - accuracy: 0.9905 - val_loss: 0.0314 - val_accuracy: 0.9909
Epoch 15/20
96/96 [==============================] - 37s 383ms/step - loss: 0.0246 - accuracy: 0.9928 - val_loss: 0.0292 - val_accuracy: 0.9922
Epoch 16/20
96/96 [==============================] - 37s 381ms/step - loss: 0.0238 - accuracy: 0.9931 - val_loss: 0.0301 - val_accuracy: 0.9909
Epoch 17/20
96/96 [==============================] - 36s 375ms/step - loss: 0.0259 - accuracy: 0.9918 - val_loss: 0.0373 - val_accuracy: 0.9883
Epoch 18/20
96/96 [==============================] - 37s 386ms/step - loss: 0.0236 - accuracy: 0.9924 - val_loss: 0.0282 - val_accuracy: 0.9909
Epoch 19/20
96/96 [==============================] - 37s 388ms/step - loss: 0.0256 - accuracy: 0.9898 - val_loss: 0.0314 - val_accuracy: 0.9896
Epoch 20/20
96/96 [==============================] - 37s 384ms/step - loss: 0.0240 - accuracy: 0.9924 - val_loss: 0.0289 - val_accuracy: 0.9922
[INFO] evaluating network...
              precision    recall  f1-score   support

   with_mask       0.99      1.00      0.99       384
without_mask       1.00      0.99      0.99       386

    accuracy                           0.99       770
   macro avg       0.99      0.99      0.99       770
weighted avg       0.99      0.99      0.99       770

[INFO] saving mask detector model...
Traceback (most recent call last):
  File "train_mask_detector.py", line 149, in <module>
    plt.plot(np.arange(0, N), H.history["acc"], label="train_acc")
KeyError: 'acc'

Limited mask detection

The code is only able to detect the mask which is of white color and the mask similar to lighter colors.
I would suggest you to use more dataset of factual images as the data set used was created by an AR program which placed white mask vector file on the images provided to it.

You should seek to this problem in order to say it a perfect model of "Face Mask Detection"

Addition to that I recommend you change the model to a compressed .h5 model

Tensorflow-lite

Any way to get the detect_mask_video.py or detect_mask_image.py working with Tensorflow lite? I don't need to train the model - I just want to run it.

Replace demo link in Readme file

Replace the YouTube demo link with: https://youtu.be/wYwW7gAYyxw