MiZhenxing
MiZhenxing
The height and width of the original image should be divisible by 32 because the depth map is 1/4 size and the 3D cost network performs "stride=2" three times.
Hi, we use the mid-point of each bin as the sample depths. So `sample_interval * (i + i + 1) / 2.0` actually computes the mid-point of i-th bin.
Thanks for your quick response. Sorry I didn't notice that example 🤣. I will do some testing on my code.
Hi, thank you for your question. `forward_one_depth` only computes the depth map of one binary tree depth. `forward_all_depth` computes the depth maps of all binary tree depths and gets the...
Hi, in our code we actually only support using model = "one" in training and only support model = "all" in testing. This is related to "Memory-efficient Training." in Paper...
Hi, sorry to be late. In Figure 2, you can see an upsample operation from Stage 2 to Stage 3 (We use "stage" to refer binary depth in the paper....
Hi, you can try the gipuma fusion in [CasMVSNet](https://github.com/alibaba/cascade-stereo/blob/master/CasMVSNet/gipuma.py)
The size of images should be divisible by 32. You could use fuctions like scale_mvs_input to change the size of images and also the intrisic parameters of your cameras.