head_detection
Dataset: https://github.com/HCIILAB/SCUT-HEAD-Dataset-Release
Trained yolo on custom for head detection.Concepts behind the yolo is mentioned in yolo_basic_working.pdf.
xml_2_txt.py converts the xml to reqiured label format i.e. for each image we need separate text file containg the details of each head on new line.The details contain the following information:
1. class_num
2. x co-ordinate of centre of the object with respect to entire image i.e.(xmin+(xmax-xmin))/image_width
3. y co-ordinate of centre of the object with respect to entire image i.e.(ymin+(ymax-ymin))/image_height
4. absolute width of the object i.e. absolute_width=(xmax-xmin)/image_width
5. absolute height of the object i.e. absolute_height=(ymax-ymin)/image_height