code_template

This project is to create a code template for image classification, object detection and segmentation. It's based on pytorch API, it supports to change the pretrained backbone on given config yaml file.

contents:

README ...................  This file
config ...................  directory contains config yaml files
core   ...................  directory contains classification/detection/segmentation part code and dataset.py
utils  ...................  directory contains some common functions and detection specific functions
data   ...................  directory contains your customerized data, it should be organized by specific format
logs   ...................  directory contains log file
models ...................  directory contains best models

environment

• python 3.x
• intel-ipex 1.2
• pytorch 1.7

training dataset

This code is verified by hymenoptera_data and pascal_voc2007

example

training

python -c config/resnet50.yml --train python -c config/resnet50.yml --train --ipex(if ipex enabled)

inference

python -c config/resnet50.yml -d path of your data python -c config/resnet50.yml -d path of your data --ipex(if ipex enabled)

future work

object detection is only support faster_rcnn topology, SSD could be integrated in this project