Convolutional neural network architecture for geometric matching
Source input image
Target input image
(Images above are from PF-WILLOW dataset. Web site url is https://www.di.ens.fr/willow/research/proposalflow/)
Output image
Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.
For the sample image and default model (streetview_affine),
$ python3 cnngeometric_pytorch.py
If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.
$ python3 cnngeometric_pytorch.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH
By adding the --video option, you can input the video.
If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.
$ python3 cnngeometric_pytorch.py --video VIDEO_PATH
If you want to change model type, use --model_1 option.
$ python3 cnngeometric_pytorch.py --model_1 MODEL_NAME
Furthermore, you can predict in two stages by adding --model_2 option.
$ python3 cnngeometric_pytorch.py --model_1 MODEL_NAME --model_2 MODEL_NAME
If you want to predict iteratively, use --num_of_iter option.
$ python3 cnngeometric_pytorch.py --num_of_iter ITER_NUM
The followings are example commands.
$ python3 cnngeometric_pytorch.py --model_1 streetview_affine --num_of_iter 3
$ python3 cnngeometric_pytorch.py --model_1 streetview_affine --model_2 streetview_tps --num_of_iter 1
$ python3 cnngeometric_pytorch.py --model_1 streetview_affine --model_2 streetview_tps --num_of_iter 2 --input input_source.png --input_tgt input_target.png --savepath sample.png
$ python3 cnngeometric_pytorch.py --model_1 streetview_affine --num_of_iter 3 --video ex1.mp4 --video_tgt ex2.mp4 --savepath output.mov
PyTorch
ONNX opset=11
streetview_affine.onnx.prototxt
streetview_hom.onnx.prototxt
streetview_tps.onnx.prototxt