CCT-Inanimate-7x224

Model Card: CCT Model for Inanimate Class Classification

Model Overview

This model is a Convolutional Contrastive Transformer (CCT) architecture trained on a dataset containing images of inanimate objects. The dataset includes non-animal classes from the ImageNet1k dataset, making the model particularly strong in recognizing non-living organism classes. The model is designed to classify images into inanimate classes, focusing on objects such as vehicles, furniture, and other non-living things.

Training Details

• Dataset: A custom dataset containing images from the inanimate object classes of the ImageNet1k dataset.
• Task: Inanimate class classification, where the model categorizes images into specific inanimate classes.
• Model Architecture: Convolutional Contrastive Transformer (CCT).
• Training Parameters: The model was trained using the parameters defined in the args.yml file that was uploaded.

Available Checkpoints

The following model checkpoints are provided:

1. model_best_sd.pth:

   • This checkpoint can be directly used for classification tasks.

2. model_best.pth.tar:

   • This checkpoint contains parameters from previous training iterations and can be used for further fine-tuning or continued training.

How to Use

1. For Classification: Use the pre-trained model available in model_best_sd.pth to directly perform image classification on inanimate object classes.

2. For Fine-Tuning: If you wish to continue training the model with your own dataset or further fine-tune it, you can start from the checkpoint model_best.pth.tar. This checkpoint includes the parameters from the previous training and allows for continued training.

Usage Instructions

# To use the model for classification:
python classify.py --model model_best_sd.pth --input <your_image_path> --output <output_path>

# To continue training:
python train.py --model model_best.pth.tar --args args.yml