whulwb / assisted_learning

The core code of "Assisted learning for land use classification: The important role of semantic correlation between heterogeneous images"

Assisted learning for land use classification: The important role of semantic correlation between heterogeneous images

This is the core code of our work published in ISPRS J. 😘

This work propose an innovative assisted learning framework that employs a "teacher-student" architecture equipped with local and global distillation schemes for land use classification on heterogeneous data.

It has several advantages as outlined below:

⭐ Ability to maintain performance during testing with missing modalities

⭐ High interpretability demonstrating knowledge transferability between different modalities

⭐ Simplicity and flexibility

Below is the process of the proposed framework:

Step 1: Training teacher models using cross-entropy and dice loss. (Using train_t.py to train the teachr.)

Step 2: Training student models using our framework. (Using train_s.py to train the teachr.)

Step 3: Testing.

---

Thanks to bubbliiiing for providing the open-source project HRnet.

assisted_learning's Issues

Large loss value in student model

Thank you for your work on this project. It's been incredibly helpful for me.

However, I've encountered a challenge that I hope to seek your advice on. During the training process of student model, I observe that the loss value turns into NaN at the first epoch, which has been quite perplexing.

Thank you in advance for your time and help.