Introduction

This repo integrates the common useful tools for object detection.

Quickstart

• Visualization

  import visualization as vis
  
  # args of vis.show
  # 1 class_list: list[str].
  # 2 img_path: str.
  # 3 ant_path: str or None.
  # 4 pd_boxes_type: '' or 'voc' or 'yolo' or 'yolo_int' or 'coco'
  # 5 pd_boxes: ndarray in shape (N,4) coco/voc/yolo
  # 6 pd_cids: ndarray in shape (N,) class index
  # 7 pd_cfs: ndarray in shape (N,) confidence
  
  # show img with coco annotation and coco prediction
  vis.show(["dog","cat"], f"data/coco/pic0.jpg", "data/coco/coco.json", "coco", pd_boxes, pd_cids, pd_cfs)
  
  # show img with voc annotation and voc prediction
  vis.show(["dog","cat"], f"data/voc/pic0.jpg", "data/voc/pic0.xml", "voc", pd_boxes, pd_cids, pd_cfs)
  
  # show img with coco annotation and coco prediction
  vis.show(["dog","cat"], f"data/coco/pic0.jpg", "data/yolo/pic0.txt", "yolo", pd_boxes, pd_cids, pd_cfs)

  See more details and examples in visualization/

• Format Conversion

  cd format_conversion

  import os
  
  import convert
  
  # convert voc to yolo
  os.makedirs("output/voc2yolo", exist_ok=True)
  convert.voc2yolo("data/voc", "output/voc2yolo", ['dog', 'cat'])
  
  # convert voc to coco
  os.makedirs("output/voc2coco", exist_ok=True)
  convert.voc2coco("data/voc", "output/voc2coco/coco.json", ['dog', 'cat'])
  
  # convert yolo to voc
  os.makedirs("output/yolo2voc", exist_ok=True)
  convert.yolo2voc("data/yolo", "output/yolo2voc", ['dog', 'cat'])
  
  # convert yolo to coco
  os.makedirs("output/yolo2coco", exist_ok=True)
  convert.yolo2coco("data/yolo", f"output/yolo2coco/coco.json", ['dog','cat'])
  
  # convert coco to voc
  os.makedirs("output/coco2voc", exist_ok=True)
  convert.coco2voc("data/coco/coco.json", "output/coco2voc")
  
  # convert coco to yolo
  os.makedirs("output/coco2yolo", exist_ok=True)
  convert.coco2yolo("data/coco/coco.json", "output/coco2yolo")

  See more details and examples in format_conversion/

Output examples

• Visualization