Open Ale0311 opened 3 years ago
@Ale0311 It is important to scale the resolution of input images to make the focal length is close to the focal length of the training images if you want use a pretrained model. For this, you can use OpenCV's remap function to simulate your images.
@Ale0311 Did you solve this problem?
May I ask what the value 518 should be replaced with if I want to use my own dataset
Hello! Could you please tell me how to scale the depth estimation if my focal length is 1101.8513?
I use the nyu pretrained model and this scaling equation:
pred_depth_scaled = pred_depth * 518.8579/ 1101.8513 * 1000
518.8579 is the focal length of the nyu dataset. However, the results are not accurate. Am I not scaling the depth in a right way?