Skip to content

Latest commit

 

History

History
41 lines (31 loc) · 1.13 KB

README.md

File metadata and controls

41 lines (31 loc) · 1.13 KB
Raw

MMdetection codes for DMPR-PS 停车位检测(parking-slot-detection)”

This is the mmdetection-implementation of DMPR-PS based on DMPR-PS

Step-1 Install mmcv 1.4.5 and MMdetection 2.19.0

Step-2 Download the dataset

Prepare the dataset and preprocess it according to DMPR-PS

Aggregate annotation files into one json, like this:

import os
import mmcv
import json
anns = list()
root = '/mnt/disk2/FedPSDet/Server/label/test/'
for file in os.listdir(root):
    if file.endswith(".json"):
        sample = {}
        sample['filename'] = os.path.splitext(file)[0]+'.jpg'
        sample['width'] = 600
        sample['height'] = 600
        sample['ann'] = mmcv.load(root+file)
        anns.append(sample)
with open('./test_psd.json', 'w') as file:
    json.dump(anns, file)

Step-3 Change the code

Overwrite our codes into the mmdetection file.

Step-4 Change the configs/FedPSD/yolo.py

Step-5 Train

bash ./tools/dist_train.sh

8张1080ti，每张卡12个样本，迭代1000次后即有0.906的ap（计算方式与DMPR完全一致）