| #use case: converting image sequence to .mp4 video | |
| #official documentation: https://trac.ffmpeg.org/wiki/Create%20a%20video%20slideshow%20from%20images | |
| #description: | |
| # An image sequence follows the pattern, defaultCanvas0-000000000.png (9 zeros) | |
| # video codec is libx264 | |
| # pixel format yuv420p | |
| # output file is named out.mp4 | |
| ffmpeg -i 'defaultCanvas0-%9d.png' -c:v libx264 -pix_fmt yuv420p out.mp4 | |
| #if you want your image sequence to start on a different number than index 0, use the -start_number parameter |
This is an Keras implementation of ResNet-152 with ImageNet pre-trained weights. I converted the weights from Caffe provided by the authors of the paper. The implementation supports both Theano and TensorFlow backends. Just in case you are curious about how the conversion is done, you can visit my blog post for more details.
ResNet Paper:
Deep Residual Learning for Image Recognition.
Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
arXiv:1512.03385