| I&T Solution |
e-Intelligent Counting on Opening of Glazed Entrance Doors in Shopping Centres of Public Rental Housing Estates
(REF: S-1319) |
| Trial Project |
|
| Solution Feature |
- Enable region proposals and feature extraction - Region Convolutional Neural Networks shares full-image convolutional features with the detection network thus enabling region proposals
- Enable a classifier to detect an image region tightly enclosing a door as a positive sample and a background region which has nothing to do with doors as a negative sample
- Enable training for the counting system based on supervised pre-training and domain-specific fine-tuning
- Enable an alert signal once the number of counts reaches the preset figure and detection of the failure of door hinge
|
| Trial Application and Expected Outcome |
- Trial conduct using R-CNN model to count the number of times a glazed entrance door is opened and closed per day over a period of time on the input images.
- Proposed regions will be selected at multiple scales with different shapes and sizes. Each region proposal will be labeled with a class and a ground-truth bounding box.
- Choose a pretrained CNN and truncate it before the output layer. The trial application will resize each region proposal to the input size required by the network, and output the extracted features for the region proposal through forward propagation.
- For extracted features (doors opening and closing), we will label bounding box of each region proposal so as to train a model to predict the ground-truth bounding box.
|
| Info on I&T Solution Provider |
| Solution Provider | : | Hexan Limited | | Address | : | Unit 607, 6/F Yen Sheng Centre, 64 Hoi Yuen Road, Kwun Tong | | Contact Person | : | Mr. HUNG |
| Position | : | System Architecture Consultant | | Tel | : | 91017517 | | Email | : |
obimlimitedhk@gmail.com |
|
