Skip to content

towserchen/ziptrak-ai

Repository files navigation

Ziptrak AI openning detection

Introduction

Use AI models to detect the positions of openings in indoor and outdoor scenes for the purpose of overlaying 3D models at those locations.

Show Doc

Author: Towser (towserchen#gmail.com)

Usage

Run Environment Requirement

Supports Windows, Linux and Mac, Python 3.10 or later must be installed. A server with a GPU/NPU is very helpful to promote the performance.

During installation, you may encounter various issues related to CUDA and torch. Please refer to the corresponding official documentation. Ideally, using a properly configured GPU that supports CUDA as the computational platform will yield the best performance.

Install

1, Clone this repo, remember to run git lfs pull to pull all weight files. If github runs out of outbound traffic, download it manually from the following address:

wget -O "weights.zip" https://tmp-hd106.vx-cdn.com/file-6722577d87b1e-6722601f0d439/weights.zip

2, Install all dependencies:

pip install -r requriements.txt

Run

Command line

1, Adjust the app.py, change variable file_path to your image that want to be detected, and is_window_detected

2, python app.py

HTTP API

Launch API server

Modify params according to your requirements and run:

uvicorn api:app --host 0.0.0.0 --port 80

Further, you need to configure yourself reverse proxy and SSL certificates if the app runs in a production environment.

API list

Upload an image and detect it

POST /detect

Params

Parameter Type Meaning Required Default
upload_file File The file to be uploaded, from HTML File Yes None
is_window_detected int Whether to detect windows, where 1 or 0 indicates True and False No 1
save_processed_images int Whether to save processed files, where 1 or 0 indicates True and False; if 1, it will save the object detection result images in the /results/ directory No 0

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published