Face Under Mask Reconstruction, Image Inpainting, and Object Detection and Segmentation with the Detectron2 library

Steps:

• Preprocessed, cleaned, and prepared custom datasets for the model with over 21,000 images🖼️ from the CelebA and UTKFace datasets with faces of different genders, ages, and races 👧🏻👩🏻👵🏻 👴🏻👨🏻👦🏻

• Trained a detector🕵🏻‍♀️ from the Detectron2 library to detect and segment the occlusions in a face🗿 image

• Built vanilla and partial convolution encoders and decoders with a discriminator for the inpainting tasks

• Trained the inpainting module separately for several hours⏳

• Achieved average accuracy🧾🏆 in the results📊