This case includes:
- The fine-tuning of the pre-trained ResNet50 model on the cat and dog classification training set,
- Evaluates the accuracy on the test set
- Builds a Demo program based on Gradio to verify the effectiveness of the trained checkpoint.
Among them, SwanLab plays a role in recording hyperparameters, tracking metrics, and visualizing dashboards in train.py.
Install the necessary libraries:
pip install -r requirements.txt
Download the dataset to the datasets folder in this directory.
Google Drive (9.9MB)｜百度云(提取码b3td)
You can start training by running
python train.py
SwanLab will track your train_loss and validation_acc, start the experiment dashboard to visualize the results by running
swanlab watch
You can start Demo based on Gradio by running
python app.py
