Seeking Advice on Optimizing Hyperparameters for Door Type Analysis in Architectural Floorplans

I’m currently tackling a challenging project that involves detecting doors in architectural floorplans. While the detection accuracy of doors is quite good, I’m facing difficulty in analyzing the door types.

There are over 25 classes of door types, such as barn_single, bifold_double, swing_single_right_hand, swing_single_left_hand, and more. Each class represents a different type of door, which adds complexity to the task.

I have a dataset of around 5000 images of floorplans, which I’m using to train the model.

I’m working on a Windows-based system, and I would greatly appreciate help with determining the hyperparameters to improve results for door type analysis. Hyperparameters are settings that control the learning process of the model, such as learning rate, batch size, and the number of training epochs.

I am using YOLOv7 as the pre-trained model.

If you’re familiar with machine learning or computer vision techniques, any advice on adjusting these hyperparameters for better results would be invaluable. Additionally, if there’s anything unclear or if you need more details, please let me know, and I’ll provide further information.

Hello! I might want to consider using an object detection model to identify the door, then crop the detection and pass it to a classification model to understand the door type.

Also, consider expanding your dataset by use flip or rotation augmentation.

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.