Downloads · 30 days
0
world-wonders-classification
sandi-irvan/world-wonders-classification
world-wonders-classification is a machine learning model from sandi-irvan. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
- Model: ResNet50 - Number of Classes: 12 - Total Epochs: 15 - Optimizer: Adam - Learning Rate: 0.001 - Loss Function: CrossEntropyLoss
Access
Public
Updated Sep 23, 2024
Repo size
94.4 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pth94.4 MB · 100%
At a glance
- Access
- Public
- Created
- Sep 23, 2024
- Updated
- Sep 23, 2024
- SHA
- 18c31174
Model card
From the Hugging Face model README
Model Training Report
Model Details
- Model: ResNet50
- Number of Classes: 12
- Total Epochs: 15
- Optimizer: Adam
- Learning Rate: 0.001
- Loss Function: CrossEntropyLoss
Training Accuracy per Epoch
| Epoch | Training Accuracy (%) |
|---|---|
| 1 | 58.32 |
| 2 | 63.15 |
| 3 | 69.45 |
| 4 | 74.21 |
| 5 | 77.58 |
| 6 | 80.32 |
| 7 | 82.01 |
| 8 | 83.50 |
| 9 | 84.72 |
| 10 | 85.30 |
| 11 | 86.85 |
| 12 | 87.20 |
| 13 | 88.10 |
| 14 | 89.25 |
| 15 | 90.30 |
Test Results
- Average Test Loss: 0.1985
- Test Accuracy: 91.76%
Additional Notes
- The model achieved 91.76% on the test set, exceeding the 90% target.
- Training improvements were consistent, with significant gains in accuracy from epoch 5 onwards.
- This was achieved using a pre-trained ResNet50 model fine-tuned on the building classification dataset.
- Further fine-tuning or dataset expansion might improve the model even more.
- the dataset that is used for training is from kaggle :https://www.kaggle.com/datasets/balabaskar/wonders-of-the-world-image-classification?resource=download """
- Created
- Sep 23, 2024
- Updated
- Sep 23, 2024