Skip to content

Latest commit

 

History

History
55 lines (42 loc) · 3 KB

README.md

File metadata and controls

55 lines (42 loc) · 3 KB

Analyze Existing Recordings and Add Bookmarks

Hope to anaylze the recording to find out when is someone loitering near your door and record the time? Or finding out the hit color of the clothes that people is wearing to decide the business? Furthermore, add bookmarks to record the time when events happened. If so, this sample is what you need! This repository demonstrates how to analyze existing recordings with your own model and add bookmarks. We develop a clothe model as an example.

The system first downloads recordings on Surveillance Station to the local computer. Then, it tracks all the people in the recordings, recognizes their wearings, and adds their wearings as bookmarks back to the recordings. Finally, users can use these bookmarks as a filter to search related recordings. In this example, you will learn:

  1. Download and analyze existing recordings from Surveillance Station
  2. How to add bookmarks to recordings.

For the detail of Clothe Model, please refer to clothe-model.pdf in this directory.

TensorFlow-2.x-YOLOv3

We use TensorFlow-2.x-YOLOv3 for objection detection and tracking. Please follow the Installation Guide in the repo to install the required dependencies.

For this project, you will only need to download the yolov3 pretrained weights into ../TensorFlow-2.x-YOLOv3/model_data.

MMFashion

We use MMFashion for clothe category classifying. Please install the required dependencies in this repo.

For this project, please download the following weights from Model Zoo in MMFashion:

  1. Download vgg16.pth in the Init models to ../mmfashion/checkpoint
  2. Download the VGG-16 Global Pooling model to ../mmfashion/checkpoint/CateAttrPredict/vgg/global

You also need to download the following files from Data Preparation in MMFashion:

  • Download list_attr_cloth.txt and list_category_cloth.txt to ../mmfashion/data/Attr_Predict/Anno_fine

Download and analyze existing recordings from Surveillance Station

After preparing the environment, you can refer to the Sample code to download and analyze existing recordings from Surveillance Station.

Run the demo

Edit ../add_recording_bookmark.py to put your NVR IP_ADDR, PORT, ACCOUNT, PASSWORD. Then, run the following command:

python add_recording_bookmark.py

Results

If you run this demo successfully, you will see the bookmarks when you open Recording.

If you click a specific recording, you can see all it's bookmarks on the right side.

You can also filter recordings with bookmarks.