Face Mask Classification Dataset Preparation
Budget: $10 – $30 USD
We're looking for a skilled developer to help prepare a high-quality dataset for a face mask classification project. This data will be used later to train model to classify whether person wearing face mask or not.
For this project scope, only data processing task you need to do, not model training.
Your Tasks:
1. Dataset Collection & Processing
- Ensure proper labeling, format consistency
Datasets:
Reallife:
1) https://www.kaggle.com/datasets/andrewmvd/face-mask-detection/data
2) https://www.kaggle.com/datasets/belsonraja/face-mask-dataset-with-and-without-mask
3) https://github.com/chandrikadeb7/Face-Mask-Detection/tree/master/dataset
4) https://data.mendeley.com/datasets/v3kry8gb59/1
5) https://www.kaggle.com/datasets/rahulmangalampalli/mafa-data
6) https://github.com/borutb-fri/FMLD
7) https://datasetninja.com/face-mask-detection
8) https://data.mendeley.com/datasets/8pn3hg99t4/2
9) https://github.com/AIZOOTech/FaceMaskDetection?tab=readme-ov-file
10) https://github.com/prajnasb/observations/tree/master/experiements/data
Synthetic:
1) https://github.com/cabani/MaskedFace-Net
2) https://github.com/X-zhangyang/Real-World-Masked-Face-Dataset
3) https://github.com/HumaticsLAB/sf-mask?tab=readme-ov-file
Additional for data generation:
1) https://www.kaggle.com/datasets/selfishgene/sfhq-t2i-synthetic-faces-from-text-2-image-models
2. Synthetic Data Generation:
- Use open-source solutions or custom scripts to programmatically overlay face masks on unmasked faces. Use 3+ methods.
1) https://github.com/aqeelanwar/MaskTheFace
2) https://github.com/securifai/masked_faces
3) Gan based any need o finalize?
Tasks:
1) Downlaod this datasets.
2) Use insightface scrfd face detector for face detection if required.
3) For each dataset. Generate cropped faces dataset with labels "no_mask", "with_mask" . Incase if dataset contains "invalid_mask" label then you can create that category as well. Create folder with this labels for each dataset and put cropped face images. Give option to specify margin around face bounding box for cropping region.
4) Test and integrate open source solution to generate new masked faces from our data.
5) Try to ensure higher level of accuracy(no misclassified labels) in processed data.
Resources Provided:
- Access to a GPU-enabled server for all data processing tasks.
For this project scope, only data processing task you need to do, not model training.
Your Tasks:
1. Dataset Collection & Processing
- Ensure proper labeling, format consistency
Datasets:
Reallife:
1) https://www.kaggle.com/datasets/andrewmvd/face-mask-detection/data
2) https://www.kaggle.com/datasets/belsonraja/face-mask-dataset-with-and-without-mask
3) https://github.com/chandrikadeb7/Face-Mask-Detection/tree/master/dataset
4) https://data.mendeley.com/datasets/v3kry8gb59/1
5) https://www.kaggle.com/datasets/rahulmangalampalli/mafa-data
6) https://github.com/borutb-fri/FMLD
7) https://datasetninja.com/face-mask-detection
8) https://data.mendeley.com/datasets/8pn3hg99t4/2
9) https://github.com/AIZOOTech/FaceMaskDetection?tab=readme-ov-file
10) https://github.com/prajnasb/observations/tree/master/experiements/data
Synthetic:
1) https://github.com/cabani/MaskedFace-Net
2) https://github.com/X-zhangyang/Real-World-Masked-Face-Dataset
3) https://github.com/HumaticsLAB/sf-mask?tab=readme-ov-file
Additional for data generation:
1) https://www.kaggle.com/datasets/selfishgene/sfhq-t2i-synthetic-faces-from-text-2-image-models
2. Synthetic Data Generation:
- Use open-source solutions or custom scripts to programmatically overlay face masks on unmasked faces. Use 3+ methods.
1) https://github.com/aqeelanwar/MaskTheFace
2) https://github.com/securifai/masked_faces
3) Gan based any need o finalize?
Tasks:
1) Downlaod this datasets.
2) Use insightface scrfd face detector for face detection if required.
3) For each dataset. Generate cropped faces dataset with labels "no_mask", "with_mask" . Incase if dataset contains "invalid_mask" label then you can create that category as well. Create folder with this labels for each dataset and put cropped face images. Give option to specify margin around face bounding box for cropping region.
4) Test and integrate open source solution to generate new masked faces from our data.
5) Try to ensure higher level of accuracy(no misclassified labels) in processed data.
Resources Provided:
- Access to a GPU-enabled server for all data processing tasks.