2_object_detection

English|中文

Detection samples

Directory structure and description

This warehouse contains a variety of detection samples for users' reference. The directory structure and specific instructions are as follows.

• VGG_SSD series samples

  Sample name	Sample description	Characteristic analysis	support chip
  VGG_SSD_coco_detection _CV_with_AIPP	Picture detection sample	Both input and output are JPG images, and opencv is used for image encoding, decoding and cropping. AIPP is used for model conversion.	Ascend310
  VGG_SSD_coco_detection _CV_without_AIPP	Picture detection sample	The input and output are all JPG images, use opencv for image encoding, decoding and cropping, and AIPP function is not used during model conversion	Ascend310
  VGG_SSD_coco_detection _DVPP_with_AIPP	Picture detection sample	The input and output are all JPG pictures, dvpp is used for picture decoding and cropping, opencv is used for picture coding, and the AIPP function is used for model conversion	Ascend310

• YOLO3 series samples

  Sample name	Sample description	Characteristic analysis	support chip
  YOLOV3_coco_detection _picture	Picture detection sample	The input and output are all JPG images, using dvpp for image decoding and cropping, using opencv for image encoding, not using the acllite public library, and the model is the yolov3-caffe network (modify the prototxt directly, and simplify the post-processing as an operator operation)	Ascend310
  YOLOV3_VOC_detection _picture	Picture detection sample	The input and output are all JPG pictures, use dvpp for picture decoding and cropping, use opencv for picture encoding, use the acllite public library, and the model is yolov3-tf network (post-processing is implemented using code)	Ascend310
  YOLOV3_coco_detection _picture_with_freetype	Picture detection sample	Input is JPG picture, output is YUV file, use dvpp for picture encoding, decoding and cropping, use acllite public library, model is yolov3-caffe network, post-processing uses freetype to write text directly on YUV picture, and use memory modification method in YUV Picture frame	Ascend310
  YOLOV3_coco_detection _dynamic_AIPP	Picture detection sample	Both input and output are JPG images, dvpp is used for image decoding and cropping, opencv is used for image encoding, acllite public library is used, the model is yolov3-caffe network, and the dynamic AIPP feature is used	Ascend310
  YOLOV3_dynamic_batch _detection_picture	Picture detection sample	The input is a BIN file, the output is screen printing, the acllite public library is not used, the model is the yolov3-caffe network, and the dynamic Batch/dynamic resolution feature is used	Ascend310
  YOLOV3_coco_detection _video	Video detection example	The input is mp4 video, and the output is the presenter interface display. Use opencv for video encoding and decoding and frame cropping, without using the acllite public library, the model is yolov3-caffe network	Ascend310
  YOLOV3_coco_detection _VENC	Video detection example	The input is mp4 video, the output is h264 file, opencv is used for video decoding and frame cropping, VENC is used for image encoding, easycl public library is not used, and the model is yolov3-caffe network	Ascend310

• Other samples

  sample	description	support chip
  YOLOV4_coco_detection _picture	Use the yolov4 model to perform predictive inference on the input picture, and print the result on the output picture.	Ascend310