/Real-Time-Face-Mask-Detection-YOLOv5

Trained Custom Model on YOLO-V5 to detect NO-Mask and play a Alarm

Primary LanguageJupyter Notebook

Real-Time-Face-Mask-Detection-Alarm

YOLOv5 architecture used for train the object detection model.

Dataset - kaggle link

This is annoted & preproccesed data that is formated in the YOLOv5 supported formated. You only have to feed the dataset.yaml file in the YOLOv5 training command. DEMO Video link

Screen.Recording.2023-05-09.at.11.55.19.AM.1.mov

Training the MODEL

Model model_link is train using 736 training images & 184 validation images not same resolution. 30 epochs are done on the dataset. The Epoch summary after running 30 epochs on the dataset:

val_batch1_labels

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss     Instances        Size
      29/29         0G    0.03029    0.02481   0.001696          92        416: 1
                 Class     Images  Instances          P           R         mAP50   
                   all        184       1042      0.891        0.794         0.87      0.467