Xclip Base Patch16 Hmdb 4 Shot

🚀 X-CLIP (base-sized model)

The X-CLIP model (base-sized, patch resolution of 16) trained in a few-shot fashion (K=4) for video classification, expanding language-image pretrained models for general video recognition.

🚀 Quick Start

The X-CLIP (base-sized model) is trained in a few-shot manner (K = 4) on HMDB - 51. It was introduced in the paper Expanding Language-Image Pretrained Models for General Video Recognition by Ni et al. and first released in this repository. This model uses 32 frames per video at a resolution of 224x224 for training.

Disclaimer: The team releasing X-CLIP did not write a model card for this model, so this model card is written by the Hugging Face team.

✨ Features

X-CLIP is a minimal extension of CLIP for general video - language understanding. The model is trained contrastively on (video, text) pairs.

This enables the model to be used for tasks such as zero - shot, few - shot, or fully supervised video classification and video - text retrieval.

📚 Documentation

Intended uses & limitations

You can use the raw model to determine how well text matches a given video. Check the model hub for fine - tuned versions on tasks that interest you.

How to use

For code examples, refer to the documentation.

Training data

This model was trained on HMDB - 51.

Preprocessing

The exact details of preprocessing during training can be found here. The exact details of preprocessing during validation can be found here.

During validation, one resizes the shorter edge of each frame and then performs center cropping to a fixed - size resolution (e.g., 224x224). Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.

Evaluation results

This model achieves a top - 1 accuracy of 57.3%.

📄 License

This model is released under the MIT license.