facebook-gad / c3d

C3D is a modified version of BVLC caffe to support 3D ConvNets.

C3D

C3D is a modified version of BVLC caffe to support 3D convolution and pooling. The main supporting features include:

• Training or fine-tuning 3D ConvNets.
  
• Extracting video features with pre-trained C3D models.
  

For more information about C3D, please refer to the C3D project website.

For general questions about Caffe, please refer to the BVLC project website for all documentation.

Updates

If you are about to start or just started your project, we would like to recommend you to use our C3D model in caffe2 link, which has better documentations and longer-term support. Additionally, we provide many more pre-trained models including R2D, R3D, MCx, rMCx, R(2+1)D, and more to come!

License

C3D is licensed under the Creative Commons Attribution-NonCommercial 3.0, as found in the LICENSE file.

Third-party Implementation:

• C3D-tensorflow with trained models by hx and XL