Macroscopic Wood Species

Image Classification

Figure 1 - Four types of wood flooring

Figure 2 - Different structures of the wood

Desciption & Background:

In South Korea, wooden flooring (나무마루) is used in a large percentage of houses. I've learned that there are 4 different types of flooring materials used. There are companies that manufacture these wood to supply the demand in interior decoration. These companies need a way to classify the same product type so that all the products won't get mixed up and more labor can be saved.

In this project, I am gonna be tackling the classification problem I mentioned above.

I decided to use CNNs models that are pre-trained on ImageNet dataset and apply transfer learning technique on the task. Choice of model is considered based on Figure 3, because we want models with good performance and fast. DesNet-201, and SE-ResNeXt50, has shown stunning results comparing to the models proposed by the paper I referenced.

The code is PyTorch based and trained in a single GTX-1080 8GBs GPU.

Keywords: Transfer Learning, CNNs, Macroscopic, Classification

Figure 3 - CNNs Models

Result Summary:

Imbalanced Training: 4 Models were evaluated.

Balanced Training:

Data augmentaion: randomly blur or sharpen minority class images
2 best models were evaluated

Code Explanation

Notebook

Reference

Silva, José Luís, et al. "Computer Vision-Based Wood Identification: A Review." Forests 13.12 (2022): 2041.
Kırbaş, İsmail, and Ahmet Çifci. "An effective and fast solution for classification of wood species: A deep transfer learning approach." Ecological Informatics 69 (2022): 101633.
Kwon, Ohkyung, et al. "Automatic wood species identification of Korean softwood based on convolutional neural networks." Journal of the Korean Wood Science and Technology 45.6 (2017): 797-808.

Page updated

Google Sites

Report abuse